ruby/load.c

Ignoring revisions in .git-blame-ignore-revs. Click here to bypass and see the normal blame view.

1454 строки
43 KiB
C
Исходник Обычный вид История

/*
* load methods from eval.c
*/
#include "dln.h"
#include "eval_intern.h"
#include "internal.h"
#include "internal/dir.h"
#include "internal/error.h"
#include "internal/file.h"
#include "internal/load.h"
#include "internal/parse.h"
#include "internal/thread.h"
#include "internal/variable.h"
#include "iseq.h"
#include "probes.h"
Rust YJIT In December 2021, we opened an [issue] to solicit feedback regarding the porting of the YJIT codebase from C99 to Rust. There were some reservations, but this project was given the go ahead by Ruby core developers and Matz. Since then, we have successfully completed the port of YJIT to Rust. The new Rust version of YJIT has reached parity with the C version, in that it passes all the CRuby tests, is able to run all of the YJIT benchmarks, and performs similarly to the C version (because it works the same way and largely generates the same machine code). We've even incorporated some design improvements, such as a more fine-grained constant invalidation mechanism which we expect will make a big difference in Ruby on Rails applications. Because we want to be careful, YJIT is guarded behind a configure option: ```shell ./configure --enable-yjit # Build YJIT in release mode ./configure --enable-yjit=dev # Build YJIT in dev/debug mode ``` By default, YJIT does not get compiled and cargo/rustc is not required. If YJIT is built in dev mode, then `cargo` is used to fetch development dependencies, but when building in release, `cargo` is not required, only `rustc`. At the moment YJIT requires Rust 1.60.0 or newer. The YJIT command-line options remain mostly unchanged, and more details about the build process are documented in `doc/yjit/yjit.md`. The CI tests have been updated and do not take any more resources than before. The development history of the Rust port is available at the following commit for interested parties: https://github.com/Shopify/ruby/commit/1fd9573d8b4b65219f1c2407f30a0a60e537f8be Our hope is that Rust YJIT will be compiled and included as a part of system packages and compiled binaries of the Ruby 3.2 release. We do not anticipate any major problems as Rust is well supported on every platform which YJIT supports, but to make sure that this process works smoothly, we would like to reach out to those who take care of building systems packages before the 3.2 release is shipped and resolve any issues that may come up. [issue]: https://bugs.ruby-lang.org/issues/18481 Co-authored-by: Maxime Chevalier-Boisvert <maximechevalierb@gmail.com> Co-authored-by: Noah Gibbs <the.codefolio.guy@gmail.com> Co-authored-by: Kevin Newton <kddnewton@gmail.com>
2022-04-19 21:40:21 +03:00
#include "darray.h"
#include "ruby/encoding.h"
#include "ruby/util.h"
static VALUE ruby_dln_librefs;
#define IS_RBEXT(e) (strcmp((e), ".rb") == 0)
#define IS_SOEXT(e) (strcmp((e), ".so") == 0 || strcmp((e), ".o") == 0)
#define IS_DLEXT(e) (strcmp((e), DLEXT) == 0)
static const char *const loadable_ext[] = {
".rb", DLEXT,
0
};
static const char *const ruby_ext[] = {
".rb",
0
};
Fix compatibility of cached expanded load path * file.c (rb_get_path_check_to_string): extract from rb_get_path_check(). We change the spec not to call to_path of String object. * file.c (rb_get_path_check_convert): extract from rb_get_path_check(). * file.c (rb_get_path_check): follow the above change. * file.c (rb_file_expand_path_fast): remove check_expand_path_args(). Instead we call it in load.c. * file.c (rb_find_file_ext_safe): use rb_get_expanded_load_path() to reduce expand cost. * file.c (rb_find_file_safe): ditto. * internal.h (rb_get_expanded_load_path): add a declaration. * internal.h (rb_get_path_check_to_string, rb_get_path_check_convert): add declarations. * load.c (rb_construct_expanded_load_path): fix for compatibility. Same checks in rb_get_path_check() are added. We don't replace $LOAD_PATH and ensure that String object of $LOAD_PATH are frozen. We don't freeze non String object and expand it every times. We add arguments for expanding load path partially and checking if load path have relative paths or non String objects. * load.c (load_path_getcwd): get current working directory for checking if it's changed when getting load path. * load.c (rb_get_expanded_load_path): fix for rebuilding cache properly. We check if current working directory is changed and rebuild expanded load path cache. We expand paths which start with ~ (User HOME) and non String objects every times for compatibility. We make this accessible from other source files. * load.c (rb_feature_provided): call rb_get_path() since we changed rb_file_expand_path_fast() not to call it. * load.c (Init_load): initialize vm->load_path_check_cache. * vm.c (rb_vm_mark): mark vm->load_path_check_cache for GC. * vm_core.h (rb_vm_struct): add vm->load_path_check_cache to store data to check load path cache validity. * test/ruby/test_require.rb (TestRequire): add tests for require compatibility related to cached expanded load path. [ruby-core:47970] [Bug #7158] git-svn-id: svn+ssh://ci.ruby-lang.org/ruby/trunk@37482 b2dd03c8-39d4-4d8f-98ff-823fe69b080e
2012-11-05 19:27:08 +04:00
enum expand_type {
EXPAND_ALL,
EXPAND_RELATIVE,
EXPAND_HOME,
EXPAND_NON_CACHE
};
/* Construct expanded load path and store it to cache.
We rebuild load path partially if the cache is invalid.
We don't cache non string object and expand it every time. We ensure that
Fix compatibility of cached expanded load path * file.c (rb_get_path_check_to_string): extract from rb_get_path_check(). We change the spec not to call to_path of String object. * file.c (rb_get_path_check_convert): extract from rb_get_path_check(). * file.c (rb_get_path_check): follow the above change. * file.c (rb_file_expand_path_fast): remove check_expand_path_args(). Instead we call it in load.c. * file.c (rb_find_file_ext_safe): use rb_get_expanded_load_path() to reduce expand cost. * file.c (rb_find_file_safe): ditto. * internal.h (rb_get_expanded_load_path): add a declaration. * internal.h (rb_get_path_check_to_string, rb_get_path_check_convert): add declarations. * load.c (rb_construct_expanded_load_path): fix for compatibility. Same checks in rb_get_path_check() are added. We don't replace $LOAD_PATH and ensure that String object of $LOAD_PATH are frozen. We don't freeze non String object and expand it every times. We add arguments for expanding load path partially and checking if load path have relative paths or non String objects. * load.c (load_path_getcwd): get current working directory for checking if it's changed when getting load path. * load.c (rb_get_expanded_load_path): fix for rebuilding cache properly. We check if current working directory is changed and rebuild expanded load path cache. We expand paths which start with ~ (User HOME) and non String objects every times for compatibility. We make this accessible from other source files. * load.c (rb_feature_provided): call rb_get_path() since we changed rb_file_expand_path_fast() not to call it. * load.c (Init_load): initialize vm->load_path_check_cache. * vm.c (rb_vm_mark): mark vm->load_path_check_cache for GC. * vm_core.h (rb_vm_struct): add vm->load_path_check_cache to store data to check load path cache validity. * test/ruby/test_require.rb (TestRequire): add tests for require compatibility related to cached expanded load path. [ruby-core:47970] [Bug #7158] git-svn-id: svn+ssh://ci.ruby-lang.org/ruby/trunk@37482 b2dd03c8-39d4-4d8f-98ff-823fe69b080e
2012-11-05 19:27:08 +04:00
string objects in $LOAD_PATH are frozen.
*/
static void
2021-09-16 19:20:10 +03:00
rb_construct_expanded_load_path(rb_vm_t *vm, enum expand_type type, int *has_relative, int *has_non_cache)
{
VALUE load_path = vm->load_path;
Fix compatibility of cached expanded load path * file.c (rb_get_path_check_to_string): extract from rb_get_path_check(). We change the spec not to call to_path of String object. * file.c (rb_get_path_check_convert): extract from rb_get_path_check(). * file.c (rb_get_path_check): follow the above change. * file.c (rb_file_expand_path_fast): remove check_expand_path_args(). Instead we call it in load.c. * file.c (rb_find_file_ext_safe): use rb_get_expanded_load_path() to reduce expand cost. * file.c (rb_find_file_safe): ditto. * internal.h (rb_get_expanded_load_path): add a declaration. * internal.h (rb_get_path_check_to_string, rb_get_path_check_convert): add declarations. * load.c (rb_construct_expanded_load_path): fix for compatibility. Same checks in rb_get_path_check() are added. We don't replace $LOAD_PATH and ensure that String object of $LOAD_PATH are frozen. We don't freeze non String object and expand it every times. We add arguments for expanding load path partially and checking if load path have relative paths or non String objects. * load.c (load_path_getcwd): get current working directory for checking if it's changed when getting load path. * load.c (rb_get_expanded_load_path): fix for rebuilding cache properly. We check if current working directory is changed and rebuild expanded load path cache. We expand paths which start with ~ (User HOME) and non String objects every times for compatibility. We make this accessible from other source files. * load.c (rb_feature_provided): call rb_get_path() since we changed rb_file_expand_path_fast() not to call it. * load.c (Init_load): initialize vm->load_path_check_cache. * vm.c (rb_vm_mark): mark vm->load_path_check_cache for GC. * vm_core.h (rb_vm_struct): add vm->load_path_check_cache to store data to check load path cache validity. * test/ruby/test_require.rb (TestRequire): add tests for require compatibility related to cached expanded load path. [ruby-core:47970] [Bug #7158] git-svn-id: svn+ssh://ci.ruby-lang.org/ruby/trunk@37482 b2dd03c8-39d4-4d8f-98ff-823fe69b080e
2012-11-05 19:27:08 +04:00
VALUE expanded_load_path = vm->expanded_load_path;
VALUE ary;
long i;
ary = rb_ary_hidden_new(RARRAY_LEN(load_path));
for (i = 0; i < RARRAY_LEN(load_path); ++i) {
VALUE path, as_str, expanded_path;
Fix compatibility of cached expanded load path * file.c (rb_get_path_check_to_string): extract from rb_get_path_check(). We change the spec not to call to_path of String object. * file.c (rb_get_path_check_convert): extract from rb_get_path_check(). * file.c (rb_get_path_check): follow the above change. * file.c (rb_file_expand_path_fast): remove check_expand_path_args(). Instead we call it in load.c. * file.c (rb_find_file_ext_safe): use rb_get_expanded_load_path() to reduce expand cost. * file.c (rb_find_file_safe): ditto. * internal.h (rb_get_expanded_load_path): add a declaration. * internal.h (rb_get_path_check_to_string, rb_get_path_check_convert): add declarations. * load.c (rb_construct_expanded_load_path): fix for compatibility. Same checks in rb_get_path_check() are added. We don't replace $LOAD_PATH and ensure that String object of $LOAD_PATH are frozen. We don't freeze non String object and expand it every times. We add arguments for expanding load path partially and checking if load path have relative paths or non String objects. * load.c (load_path_getcwd): get current working directory for checking if it's changed when getting load path. * load.c (rb_get_expanded_load_path): fix for rebuilding cache properly. We check if current working directory is changed and rebuild expanded load path cache. We expand paths which start with ~ (User HOME) and non String objects every times for compatibility. We make this accessible from other source files. * load.c (rb_feature_provided): call rb_get_path() since we changed rb_file_expand_path_fast() not to call it. * load.c (Init_load): initialize vm->load_path_check_cache. * vm.c (rb_vm_mark): mark vm->load_path_check_cache for GC. * vm_core.h (rb_vm_struct): add vm->load_path_check_cache to store data to check load path cache validity. * test/ruby/test_require.rb (TestRequire): add tests for require compatibility related to cached expanded load path. [ruby-core:47970] [Bug #7158] git-svn-id: svn+ssh://ci.ruby-lang.org/ruby/trunk@37482 b2dd03c8-39d4-4d8f-98ff-823fe69b080e
2012-11-05 19:27:08 +04:00
int is_string, non_cache;
char *as_cstr;
as_str = path = RARRAY_AREF(load_path, i);
Fix compatibility of cached expanded load path * file.c (rb_get_path_check_to_string): extract from rb_get_path_check(). We change the spec not to call to_path of String object. * file.c (rb_get_path_check_convert): extract from rb_get_path_check(). * file.c (rb_get_path_check): follow the above change. * file.c (rb_file_expand_path_fast): remove check_expand_path_args(). Instead we call it in load.c. * file.c (rb_find_file_ext_safe): use rb_get_expanded_load_path() to reduce expand cost. * file.c (rb_find_file_safe): ditto. * internal.h (rb_get_expanded_load_path): add a declaration. * internal.h (rb_get_path_check_to_string, rb_get_path_check_convert): add declarations. * load.c (rb_construct_expanded_load_path): fix for compatibility. Same checks in rb_get_path_check() are added. We don't replace $LOAD_PATH and ensure that String object of $LOAD_PATH are frozen. We don't freeze non String object and expand it every times. We add arguments for expanding load path partially and checking if load path have relative paths or non String objects. * load.c (load_path_getcwd): get current working directory for checking if it's changed when getting load path. * load.c (rb_get_expanded_load_path): fix for rebuilding cache properly. We check if current working directory is changed and rebuild expanded load path cache. We expand paths which start with ~ (User HOME) and non String objects every times for compatibility. We make this accessible from other source files. * load.c (rb_feature_provided): call rb_get_path() since we changed rb_file_expand_path_fast() not to call it. * load.c (Init_load): initialize vm->load_path_check_cache. * vm.c (rb_vm_mark): mark vm->load_path_check_cache for GC. * vm_core.h (rb_vm_struct): add vm->load_path_check_cache to store data to check load path cache validity. * test/ruby/test_require.rb (TestRequire): add tests for require compatibility related to cached expanded load path. [ruby-core:47970] [Bug #7158] git-svn-id: svn+ssh://ci.ruby-lang.org/ruby/trunk@37482 b2dd03c8-39d4-4d8f-98ff-823fe69b080e
2012-11-05 19:27:08 +04:00
is_string = RB_TYPE_P(path, T_STRING) ? 1 : 0;
non_cache = !is_string ? 1 : 0;
as_str = rb_get_path_check_to_string(path);
Fix compatibility of cached expanded load path * file.c (rb_get_path_check_to_string): extract from rb_get_path_check(). We change the spec not to call to_path of String object. * file.c (rb_get_path_check_convert): extract from rb_get_path_check(). * file.c (rb_get_path_check): follow the above change. * file.c (rb_file_expand_path_fast): remove check_expand_path_args(). Instead we call it in load.c. * file.c (rb_find_file_ext_safe): use rb_get_expanded_load_path() to reduce expand cost. * file.c (rb_find_file_safe): ditto. * internal.h (rb_get_expanded_load_path): add a declaration. * internal.h (rb_get_path_check_to_string, rb_get_path_check_convert): add declarations. * load.c (rb_construct_expanded_load_path): fix for compatibility. Same checks in rb_get_path_check() are added. We don't replace $LOAD_PATH and ensure that String object of $LOAD_PATH are frozen. We don't freeze non String object and expand it every times. We add arguments for expanding load path partially and checking if load path have relative paths or non String objects. * load.c (load_path_getcwd): get current working directory for checking if it's changed when getting load path. * load.c (rb_get_expanded_load_path): fix for rebuilding cache properly. We check if current working directory is changed and rebuild expanded load path cache. We expand paths which start with ~ (User HOME) and non String objects every times for compatibility. We make this accessible from other source files. * load.c (rb_feature_provided): call rb_get_path() since we changed rb_file_expand_path_fast() not to call it. * load.c (Init_load): initialize vm->load_path_check_cache. * vm.c (rb_vm_mark): mark vm->load_path_check_cache for GC. * vm_core.h (rb_vm_struct): add vm->load_path_check_cache to store data to check load path cache validity. * test/ruby/test_require.rb (TestRequire): add tests for require compatibility related to cached expanded load path. [ruby-core:47970] [Bug #7158] git-svn-id: svn+ssh://ci.ruby-lang.org/ruby/trunk@37482 b2dd03c8-39d4-4d8f-98ff-823fe69b080e
2012-11-05 19:27:08 +04:00
as_cstr = RSTRING_PTR(as_str);
2022-07-21 19:23:58 +03:00
Fix compatibility of cached expanded load path * file.c (rb_get_path_check_to_string): extract from rb_get_path_check(). We change the spec not to call to_path of String object. * file.c (rb_get_path_check_convert): extract from rb_get_path_check(). * file.c (rb_get_path_check): follow the above change. * file.c (rb_file_expand_path_fast): remove check_expand_path_args(). Instead we call it in load.c. * file.c (rb_find_file_ext_safe): use rb_get_expanded_load_path() to reduce expand cost. * file.c (rb_find_file_safe): ditto. * internal.h (rb_get_expanded_load_path): add a declaration. * internal.h (rb_get_path_check_to_string, rb_get_path_check_convert): add declarations. * load.c (rb_construct_expanded_load_path): fix for compatibility. Same checks in rb_get_path_check() are added. We don't replace $LOAD_PATH and ensure that String object of $LOAD_PATH are frozen. We don't freeze non String object and expand it every times. We add arguments for expanding load path partially and checking if load path have relative paths or non String objects. * load.c (load_path_getcwd): get current working directory for checking if it's changed when getting load path. * load.c (rb_get_expanded_load_path): fix for rebuilding cache properly. We check if current working directory is changed and rebuild expanded load path cache. We expand paths which start with ~ (User HOME) and non String objects every times for compatibility. We make this accessible from other source files. * load.c (rb_feature_provided): call rb_get_path() since we changed rb_file_expand_path_fast() not to call it. * load.c (Init_load): initialize vm->load_path_check_cache. * vm.c (rb_vm_mark): mark vm->load_path_check_cache for GC. * vm_core.h (rb_vm_struct): add vm->load_path_check_cache to store data to check load path cache validity. * test/ruby/test_require.rb (TestRequire): add tests for require compatibility related to cached expanded load path. [ruby-core:47970] [Bug #7158] git-svn-id: svn+ssh://ci.ruby-lang.org/ruby/trunk@37482 b2dd03c8-39d4-4d8f-98ff-823fe69b080e
2012-11-05 19:27:08 +04:00
if (!non_cache) {
if ((type == EXPAND_RELATIVE &&
rb_is_absolute_path(as_cstr)) ||
(type == EXPAND_HOME &&
(!as_cstr[0] || as_cstr[0] != '~')) ||
(type == EXPAND_NON_CACHE)) {
/* Use cached expanded path. */
rb_ary_push(ary, RARRAY_AREF(expanded_load_path, i));
Fix compatibility of cached expanded load path * file.c (rb_get_path_check_to_string): extract from rb_get_path_check(). We change the spec not to call to_path of String object. * file.c (rb_get_path_check_convert): extract from rb_get_path_check(). * file.c (rb_get_path_check): follow the above change. * file.c (rb_file_expand_path_fast): remove check_expand_path_args(). Instead we call it in load.c. * file.c (rb_find_file_ext_safe): use rb_get_expanded_load_path() to reduce expand cost. * file.c (rb_find_file_safe): ditto. * internal.h (rb_get_expanded_load_path): add a declaration. * internal.h (rb_get_path_check_to_string, rb_get_path_check_convert): add declarations. * load.c (rb_construct_expanded_load_path): fix for compatibility. Same checks in rb_get_path_check() are added. We don't replace $LOAD_PATH and ensure that String object of $LOAD_PATH are frozen. We don't freeze non String object and expand it every times. We add arguments for expanding load path partially and checking if load path have relative paths or non String objects. * load.c (load_path_getcwd): get current working directory for checking if it's changed when getting load path. * load.c (rb_get_expanded_load_path): fix for rebuilding cache properly. We check if current working directory is changed and rebuild expanded load path cache. We expand paths which start with ~ (User HOME) and non String objects every times for compatibility. We make this accessible from other source files. * load.c (rb_feature_provided): call rb_get_path() since we changed rb_file_expand_path_fast() not to call it. * load.c (Init_load): initialize vm->load_path_check_cache. * vm.c (rb_vm_mark): mark vm->load_path_check_cache for GC. * vm_core.h (rb_vm_struct): add vm->load_path_check_cache to store data to check load path cache validity. * test/ruby/test_require.rb (TestRequire): add tests for require compatibility related to cached expanded load path. [ruby-core:47970] [Bug #7158] git-svn-id: svn+ssh://ci.ruby-lang.org/ruby/trunk@37482 b2dd03c8-39d4-4d8f-98ff-823fe69b080e
2012-11-05 19:27:08 +04:00
continue;
}
}
if (!*has_relative && !rb_is_absolute_path(as_cstr))
*has_relative = 1;
if (!*has_non_cache && non_cache)
*has_non_cache = 1;
/* Freeze only string object. We expand other objects every time. */
Fix compatibility of cached expanded load path * file.c (rb_get_path_check_to_string): extract from rb_get_path_check(). We change the spec not to call to_path of String object. * file.c (rb_get_path_check_convert): extract from rb_get_path_check(). * file.c (rb_get_path_check): follow the above change. * file.c (rb_file_expand_path_fast): remove check_expand_path_args(). Instead we call it in load.c. * file.c (rb_find_file_ext_safe): use rb_get_expanded_load_path() to reduce expand cost. * file.c (rb_find_file_safe): ditto. * internal.h (rb_get_expanded_load_path): add a declaration. * internal.h (rb_get_path_check_to_string, rb_get_path_check_convert): add declarations. * load.c (rb_construct_expanded_load_path): fix for compatibility. Same checks in rb_get_path_check() are added. We don't replace $LOAD_PATH and ensure that String object of $LOAD_PATH are frozen. We don't freeze non String object and expand it every times. We add arguments for expanding load path partially and checking if load path have relative paths or non String objects. * load.c (load_path_getcwd): get current working directory for checking if it's changed when getting load path. * load.c (rb_get_expanded_load_path): fix for rebuilding cache properly. We check if current working directory is changed and rebuild expanded load path cache. We expand paths which start with ~ (User HOME) and non String objects every times for compatibility. We make this accessible from other source files. * load.c (rb_feature_provided): call rb_get_path() since we changed rb_file_expand_path_fast() not to call it. * load.c (Init_load): initialize vm->load_path_check_cache. * vm.c (rb_vm_mark): mark vm->load_path_check_cache for GC. * vm_core.h (rb_vm_struct): add vm->load_path_check_cache to store data to check load path cache validity. * test/ruby/test_require.rb (TestRequire): add tests for require compatibility related to cached expanded load path. [ruby-core:47970] [Bug #7158] git-svn-id: svn+ssh://ci.ruby-lang.org/ruby/trunk@37482 b2dd03c8-39d4-4d8f-98ff-823fe69b080e
2012-11-05 19:27:08 +04:00
if (is_string)
rb_str_freeze(path);
as_str = rb_get_path_check_convert(as_str);
expanded_path = rb_check_realpath(Qnil, as_str, NULL);
if (NIL_P(expanded_path)) expanded_path = as_str;
rb_ary_push(ary, rb_fstring(expanded_path));
}
rb_obj_freeze(ary);
vm->expanded_load_path = ary;
rb_ary_replace(vm->load_path_snapshot, vm->load_path);
}
2021-09-16 19:20:10 +03:00
static VALUE
get_expanded_load_path(rb_vm_t *vm)
{
Fix compatibility of cached expanded load path * file.c (rb_get_path_check_to_string): extract from rb_get_path_check(). We change the spec not to call to_path of String object. * file.c (rb_get_path_check_convert): extract from rb_get_path_check(). * file.c (rb_get_path_check): follow the above change. * file.c (rb_file_expand_path_fast): remove check_expand_path_args(). Instead we call it in load.c. * file.c (rb_find_file_ext_safe): use rb_get_expanded_load_path() to reduce expand cost. * file.c (rb_find_file_safe): ditto. * internal.h (rb_get_expanded_load_path): add a declaration. * internal.h (rb_get_path_check_to_string, rb_get_path_check_convert): add declarations. * load.c (rb_construct_expanded_load_path): fix for compatibility. Same checks in rb_get_path_check() are added. We don't replace $LOAD_PATH and ensure that String object of $LOAD_PATH are frozen. We don't freeze non String object and expand it every times. We add arguments for expanding load path partially and checking if load path have relative paths or non String objects. * load.c (load_path_getcwd): get current working directory for checking if it's changed when getting load path. * load.c (rb_get_expanded_load_path): fix for rebuilding cache properly. We check if current working directory is changed and rebuild expanded load path cache. We expand paths which start with ~ (User HOME) and non String objects every times for compatibility. We make this accessible from other source files. * load.c (rb_feature_provided): call rb_get_path() since we changed rb_file_expand_path_fast() not to call it. * load.c (Init_load): initialize vm->load_path_check_cache. * vm.c (rb_vm_mark): mark vm->load_path_check_cache for GC. * vm_core.h (rb_vm_struct): add vm->load_path_check_cache to store data to check load path cache validity. * test/ruby/test_require.rb (TestRequire): add tests for require compatibility related to cached expanded load path. [ruby-core:47970] [Bug #7158] git-svn-id: svn+ssh://ci.ruby-lang.org/ruby/trunk@37482 b2dd03c8-39d4-4d8f-98ff-823fe69b080e
2012-11-05 19:27:08 +04:00
const VALUE non_cache = Qtrue;
if (!rb_ary_shared_with_p(vm->load_path_snapshot, vm->load_path)) {
Fix compatibility of cached expanded load path * file.c (rb_get_path_check_to_string): extract from rb_get_path_check(). We change the spec not to call to_path of String object. * file.c (rb_get_path_check_convert): extract from rb_get_path_check(). * file.c (rb_get_path_check): follow the above change. * file.c (rb_file_expand_path_fast): remove check_expand_path_args(). Instead we call it in load.c. * file.c (rb_find_file_ext_safe): use rb_get_expanded_load_path() to reduce expand cost. * file.c (rb_find_file_safe): ditto. * internal.h (rb_get_expanded_load_path): add a declaration. * internal.h (rb_get_path_check_to_string, rb_get_path_check_convert): add declarations. * load.c (rb_construct_expanded_load_path): fix for compatibility. Same checks in rb_get_path_check() are added. We don't replace $LOAD_PATH and ensure that String object of $LOAD_PATH are frozen. We don't freeze non String object and expand it every times. We add arguments for expanding load path partially and checking if load path have relative paths or non String objects. * load.c (load_path_getcwd): get current working directory for checking if it's changed when getting load path. * load.c (rb_get_expanded_load_path): fix for rebuilding cache properly. We check if current working directory is changed and rebuild expanded load path cache. We expand paths which start with ~ (User HOME) and non String objects every times for compatibility. We make this accessible from other source files. * load.c (rb_feature_provided): call rb_get_path() since we changed rb_file_expand_path_fast() not to call it. * load.c (Init_load): initialize vm->load_path_check_cache. * vm.c (rb_vm_mark): mark vm->load_path_check_cache for GC. * vm_core.h (rb_vm_struct): add vm->load_path_check_cache to store data to check load path cache validity. * test/ruby/test_require.rb (TestRequire): add tests for require compatibility related to cached expanded load path. [ruby-core:47970] [Bug #7158] git-svn-id: svn+ssh://ci.ruby-lang.org/ruby/trunk@37482 b2dd03c8-39d4-4d8f-98ff-823fe69b080e
2012-11-05 19:27:08 +04:00
/* The load path was modified. Rebuild the expanded load path. */
int has_relative = 0, has_non_cache = 0;
2021-09-16 19:20:10 +03:00
rb_construct_expanded_load_path(vm, EXPAND_ALL, &has_relative, &has_non_cache);
Fix compatibility of cached expanded load path * file.c (rb_get_path_check_to_string): extract from rb_get_path_check(). We change the spec not to call to_path of String object. * file.c (rb_get_path_check_convert): extract from rb_get_path_check(). * file.c (rb_get_path_check): follow the above change. * file.c (rb_file_expand_path_fast): remove check_expand_path_args(). Instead we call it in load.c. * file.c (rb_find_file_ext_safe): use rb_get_expanded_load_path() to reduce expand cost. * file.c (rb_find_file_safe): ditto. * internal.h (rb_get_expanded_load_path): add a declaration. * internal.h (rb_get_path_check_to_string, rb_get_path_check_convert): add declarations. * load.c (rb_construct_expanded_load_path): fix for compatibility. Same checks in rb_get_path_check() are added. We don't replace $LOAD_PATH and ensure that String object of $LOAD_PATH are frozen. We don't freeze non String object and expand it every times. We add arguments for expanding load path partially and checking if load path have relative paths or non String objects. * load.c (load_path_getcwd): get current working directory for checking if it's changed when getting load path. * load.c (rb_get_expanded_load_path): fix for rebuilding cache properly. We check if current working directory is changed and rebuild expanded load path cache. We expand paths which start with ~ (User HOME) and non String objects every times for compatibility. We make this accessible from other source files. * load.c (rb_feature_provided): call rb_get_path() since we changed rb_file_expand_path_fast() not to call it. * load.c (Init_load): initialize vm->load_path_check_cache. * vm.c (rb_vm_mark): mark vm->load_path_check_cache for GC. * vm_core.h (rb_vm_struct): add vm->load_path_check_cache to store data to check load path cache validity. * test/ruby/test_require.rb (TestRequire): add tests for require compatibility related to cached expanded load path. [ruby-core:47970] [Bug #7158] git-svn-id: svn+ssh://ci.ruby-lang.org/ruby/trunk@37482 b2dd03c8-39d4-4d8f-98ff-823fe69b080e
2012-11-05 19:27:08 +04:00
if (has_relative) {
vm->load_path_check_cache = rb_dir_getwd_ospath();
Fix compatibility of cached expanded load path * file.c (rb_get_path_check_to_string): extract from rb_get_path_check(). We change the spec not to call to_path of String object. * file.c (rb_get_path_check_convert): extract from rb_get_path_check(). * file.c (rb_get_path_check): follow the above change. * file.c (rb_file_expand_path_fast): remove check_expand_path_args(). Instead we call it in load.c. * file.c (rb_find_file_ext_safe): use rb_get_expanded_load_path() to reduce expand cost. * file.c (rb_find_file_safe): ditto. * internal.h (rb_get_expanded_load_path): add a declaration. * internal.h (rb_get_path_check_to_string, rb_get_path_check_convert): add declarations. * load.c (rb_construct_expanded_load_path): fix for compatibility. Same checks in rb_get_path_check() are added. We don't replace $LOAD_PATH and ensure that String object of $LOAD_PATH are frozen. We don't freeze non String object and expand it every times. We add arguments for expanding load path partially and checking if load path have relative paths or non String objects. * load.c (load_path_getcwd): get current working directory for checking if it's changed when getting load path. * load.c (rb_get_expanded_load_path): fix for rebuilding cache properly. We check if current working directory is changed and rebuild expanded load path cache. We expand paths which start with ~ (User HOME) and non String objects every times for compatibility. We make this accessible from other source files. * load.c (rb_feature_provided): call rb_get_path() since we changed rb_file_expand_path_fast() not to call it. * load.c (Init_load): initialize vm->load_path_check_cache. * vm.c (rb_vm_mark): mark vm->load_path_check_cache for GC. * vm_core.h (rb_vm_struct): add vm->load_path_check_cache to store data to check load path cache validity. * test/ruby/test_require.rb (TestRequire): add tests for require compatibility related to cached expanded load path. [ruby-core:47970] [Bug #7158] git-svn-id: svn+ssh://ci.ruby-lang.org/ruby/trunk@37482 b2dd03c8-39d4-4d8f-98ff-823fe69b080e
2012-11-05 19:27:08 +04:00
}
else if (has_non_cache) {
/* Non string object. */
vm->load_path_check_cache = non_cache;
}
else {
vm->load_path_check_cache = 0;
}
}
else if (vm->load_path_check_cache == non_cache) {
int has_relative = 1, has_non_cache = 1;
/* Expand only non-cacheable objects. */
2021-09-16 19:20:10 +03:00
rb_construct_expanded_load_path(vm, EXPAND_NON_CACHE,
Fix compatibility of cached expanded load path * file.c (rb_get_path_check_to_string): extract from rb_get_path_check(). We change the spec not to call to_path of String object. * file.c (rb_get_path_check_convert): extract from rb_get_path_check(). * file.c (rb_get_path_check): follow the above change. * file.c (rb_file_expand_path_fast): remove check_expand_path_args(). Instead we call it in load.c. * file.c (rb_find_file_ext_safe): use rb_get_expanded_load_path() to reduce expand cost. * file.c (rb_find_file_safe): ditto. * internal.h (rb_get_expanded_load_path): add a declaration. * internal.h (rb_get_path_check_to_string, rb_get_path_check_convert): add declarations. * load.c (rb_construct_expanded_load_path): fix for compatibility. Same checks in rb_get_path_check() are added. We don't replace $LOAD_PATH and ensure that String object of $LOAD_PATH are frozen. We don't freeze non String object and expand it every times. We add arguments for expanding load path partially and checking if load path have relative paths or non String objects. * load.c (load_path_getcwd): get current working directory for checking if it's changed when getting load path. * load.c (rb_get_expanded_load_path): fix for rebuilding cache properly. We check if current working directory is changed and rebuild expanded load path cache. We expand paths which start with ~ (User HOME) and non String objects every times for compatibility. We make this accessible from other source files. * load.c (rb_feature_provided): call rb_get_path() since we changed rb_file_expand_path_fast() not to call it. * load.c (Init_load): initialize vm->load_path_check_cache. * vm.c (rb_vm_mark): mark vm->load_path_check_cache for GC. * vm_core.h (rb_vm_struct): add vm->load_path_check_cache to store data to check load path cache validity. * test/ruby/test_require.rb (TestRequire): add tests for require compatibility related to cached expanded load path. [ruby-core:47970] [Bug #7158] git-svn-id: svn+ssh://ci.ruby-lang.org/ruby/trunk@37482 b2dd03c8-39d4-4d8f-98ff-823fe69b080e
2012-11-05 19:27:08 +04:00
&has_relative, &has_non_cache);
}
else if (vm->load_path_check_cache) {
int has_relative = 1, has_non_cache = 1;
VALUE cwd = rb_dir_getwd_ospath();
Fix compatibility of cached expanded load path * file.c (rb_get_path_check_to_string): extract from rb_get_path_check(). We change the spec not to call to_path of String object. * file.c (rb_get_path_check_convert): extract from rb_get_path_check(). * file.c (rb_get_path_check): follow the above change. * file.c (rb_file_expand_path_fast): remove check_expand_path_args(). Instead we call it in load.c. * file.c (rb_find_file_ext_safe): use rb_get_expanded_load_path() to reduce expand cost. * file.c (rb_find_file_safe): ditto. * internal.h (rb_get_expanded_load_path): add a declaration. * internal.h (rb_get_path_check_to_string, rb_get_path_check_convert): add declarations. * load.c (rb_construct_expanded_load_path): fix for compatibility. Same checks in rb_get_path_check() are added. We don't replace $LOAD_PATH and ensure that String object of $LOAD_PATH are frozen. We don't freeze non String object and expand it every times. We add arguments for expanding load path partially and checking if load path have relative paths or non String objects. * load.c (load_path_getcwd): get current working directory for checking if it's changed when getting load path. * load.c (rb_get_expanded_load_path): fix for rebuilding cache properly. We check if current working directory is changed and rebuild expanded load path cache. We expand paths which start with ~ (User HOME) and non String objects every times for compatibility. We make this accessible from other source files. * load.c (rb_feature_provided): call rb_get_path() since we changed rb_file_expand_path_fast() not to call it. * load.c (Init_load): initialize vm->load_path_check_cache. * vm.c (rb_vm_mark): mark vm->load_path_check_cache for GC. * vm_core.h (rb_vm_struct): add vm->load_path_check_cache to store data to check load path cache validity. * test/ruby/test_require.rb (TestRequire): add tests for require compatibility related to cached expanded load path. [ruby-core:47970] [Bug #7158] git-svn-id: svn+ssh://ci.ruby-lang.org/ruby/trunk@37482 b2dd03c8-39d4-4d8f-98ff-823fe69b080e
2012-11-05 19:27:08 +04:00
if (!rb_str_equal(vm->load_path_check_cache, cwd)) {
/* Current working directory or filesystem encoding was changed.
Expand relative load path and non-cacheable objects again. */
vm->load_path_check_cache = cwd;
2021-09-16 19:20:10 +03:00
rb_construct_expanded_load_path(vm, EXPAND_RELATIVE,
Fix compatibility of cached expanded load path * file.c (rb_get_path_check_to_string): extract from rb_get_path_check(). We change the spec not to call to_path of String object. * file.c (rb_get_path_check_convert): extract from rb_get_path_check(). * file.c (rb_get_path_check): follow the above change. * file.c (rb_file_expand_path_fast): remove check_expand_path_args(). Instead we call it in load.c. * file.c (rb_find_file_ext_safe): use rb_get_expanded_load_path() to reduce expand cost. * file.c (rb_find_file_safe): ditto. * internal.h (rb_get_expanded_load_path): add a declaration. * internal.h (rb_get_path_check_to_string, rb_get_path_check_convert): add declarations. * load.c (rb_construct_expanded_load_path): fix for compatibility. Same checks in rb_get_path_check() are added. We don't replace $LOAD_PATH and ensure that String object of $LOAD_PATH are frozen. We don't freeze non String object and expand it every times. We add arguments for expanding load path partially and checking if load path have relative paths or non String objects. * load.c (load_path_getcwd): get current working directory for checking if it's changed when getting load path. * load.c (rb_get_expanded_load_path): fix for rebuilding cache properly. We check if current working directory is changed and rebuild expanded load path cache. We expand paths which start with ~ (User HOME) and non String objects every times for compatibility. We make this accessible from other source files. * load.c (rb_feature_provided): call rb_get_path() since we changed rb_file_expand_path_fast() not to call it. * load.c (Init_load): initialize vm->load_path_check_cache. * vm.c (rb_vm_mark): mark vm->load_path_check_cache for GC. * vm_core.h (rb_vm_struct): add vm->load_path_check_cache to store data to check load path cache validity. * test/ruby/test_require.rb (TestRequire): add tests for require compatibility related to cached expanded load path. [ruby-core:47970] [Bug #7158] git-svn-id: svn+ssh://ci.ruby-lang.org/ruby/trunk@37482 b2dd03c8-39d4-4d8f-98ff-823fe69b080e
2012-11-05 19:27:08 +04:00
&has_relative, &has_non_cache);
}
else {
/* Expand only tilde (User HOME) and non-cacheable objects. */
2021-09-16 19:20:10 +03:00
rb_construct_expanded_load_path(vm, EXPAND_HOME,
Fix compatibility of cached expanded load path * file.c (rb_get_path_check_to_string): extract from rb_get_path_check(). We change the spec not to call to_path of String object. * file.c (rb_get_path_check_convert): extract from rb_get_path_check(). * file.c (rb_get_path_check): follow the above change. * file.c (rb_file_expand_path_fast): remove check_expand_path_args(). Instead we call it in load.c. * file.c (rb_find_file_ext_safe): use rb_get_expanded_load_path() to reduce expand cost. * file.c (rb_find_file_safe): ditto. * internal.h (rb_get_expanded_load_path): add a declaration. * internal.h (rb_get_path_check_to_string, rb_get_path_check_convert): add declarations. * load.c (rb_construct_expanded_load_path): fix for compatibility. Same checks in rb_get_path_check() are added. We don't replace $LOAD_PATH and ensure that String object of $LOAD_PATH are frozen. We don't freeze non String object and expand it every times. We add arguments for expanding load path partially and checking if load path have relative paths or non String objects. * load.c (load_path_getcwd): get current working directory for checking if it's changed when getting load path. * load.c (rb_get_expanded_load_path): fix for rebuilding cache properly. We check if current working directory is changed and rebuild expanded load path cache. We expand paths which start with ~ (User HOME) and non String objects every times for compatibility. We make this accessible from other source files. * load.c (rb_feature_provided): call rb_get_path() since we changed rb_file_expand_path_fast() not to call it. * load.c (Init_load): initialize vm->load_path_check_cache. * vm.c (rb_vm_mark): mark vm->load_path_check_cache for GC. * vm_core.h (rb_vm_struct): add vm->load_path_check_cache to store data to check load path cache validity. * test/ruby/test_require.rb (TestRequire): add tests for require compatibility related to cached expanded load path. [ruby-core:47970] [Bug #7158] git-svn-id: svn+ssh://ci.ruby-lang.org/ruby/trunk@37482 b2dd03c8-39d4-4d8f-98ff-823fe69b080e
2012-11-05 19:27:08 +04:00
&has_relative, &has_non_cache);
}
}
return vm->expanded_load_path;
}
2021-09-16 19:20:10 +03:00
VALUE
rb_get_expanded_load_path(void)
{
return get_expanded_load_path(GET_VM());
}
static VALUE
load_path_getter(ID id, VALUE * p)
{
rb_vm_t *vm = (void *)p;
return vm->load_path;
}
static VALUE
2021-09-16 19:20:10 +03:00
get_loaded_features(rb_vm_t *vm)
{
2021-09-16 19:20:10 +03:00
return vm->loaded_features;
}
Do not load file with same realpath twice when requiring This fixes issues with paths being loaded twice in certain cases when symlinks are used. It took me multiple attempts to get this working. My original attempt tried to convert paths to realpaths before adding them to $LOADED_FEATURES. Unfortunately, this doesn't work well with the loaded feature index, which is based off load paths and not realpaths. While I was able to get require working, I'm fairly sure the loaded feature index was not being used as expected, which would have significant performance implications. Additionally, I was never able to get that approach working with autoload when autoloading a non-realpath file. It also broke some specs. This takes a more conservative approach. Directly before loading the file, if the file with the same realpath has been required, the loading of the file is skipped. The realpaths are stored as fstrings in a hidden hash. When rebuilding the loaded feature index, the hash of realpaths is also rebuilt. I'm guessing this makes rebuilding process slower, but I don think that is a hot path. In general, modifying loaded features is only done when reloading, and that tends to be in non-production environments. Change test_require_with_loaded_features_pop test to use 30 threads and 300 iterations, instead of 4 threads and 1000 iterations. I saw only sporadic failures with 4/1000, but consistent failures 30/300 threads. These failures were due to the fact that the concurrent deletions from $LOADED_FEATURES in other threads can result in rb_ary_entry returning nil when rebuilding the loaded features index. To avoid concurrency issues when rebuilding the loaded features index, the building of the index itself is left alone, and afterwards, a separate loop is done on a copy of the loaded feature snapshot in order to rebuild the realpaths hash. Fixes [Bug #17885]
2021-06-30 23:50:19 +03:00
static VALUE
2021-09-16 19:20:10 +03:00
get_loaded_features_realpaths(rb_vm_t *vm)
Do not load file with same realpath twice when requiring This fixes issues with paths being loaded twice in certain cases when symlinks are used. It took me multiple attempts to get this working. My original attempt tried to convert paths to realpaths before adding them to $LOADED_FEATURES. Unfortunately, this doesn't work well with the loaded feature index, which is based off load paths and not realpaths. While I was able to get require working, I'm fairly sure the loaded feature index was not being used as expected, which would have significant performance implications. Additionally, I was never able to get that approach working with autoload when autoloading a non-realpath file. It also broke some specs. This takes a more conservative approach. Directly before loading the file, if the file with the same realpath has been required, the loading of the file is skipped. The realpaths are stored as fstrings in a hidden hash. When rebuilding the loaded feature index, the hash of realpaths is also rebuilt. I'm guessing this makes rebuilding process slower, but I don think that is a hot path. In general, modifying loaded features is only done when reloading, and that tends to be in non-production environments. Change test_require_with_loaded_features_pop test to use 30 threads and 300 iterations, instead of 4 threads and 1000 iterations. I saw only sporadic failures with 4/1000, but consistent failures 30/300 threads. These failures were due to the fact that the concurrent deletions from $LOADED_FEATURES in other threads can result in rb_ary_entry returning nil when rebuilding the loaded features index. To avoid concurrency issues when rebuilding the loaded features index, the building of the index itself is left alone, and afterwards, a separate loop is done on a copy of the loaded feature snapshot in order to rebuild the realpaths hash. Fixes [Bug #17885]
2021-06-30 23:50:19 +03:00
{
2021-09-16 19:20:10 +03:00
return vm->loaded_features_realpaths;
Do not load file with same realpath twice when requiring This fixes issues with paths being loaded twice in certain cases when symlinks are used. It took me multiple attempts to get this working. My original attempt tried to convert paths to realpaths before adding them to $LOADED_FEATURES. Unfortunately, this doesn't work well with the loaded feature index, which is based off load paths and not realpaths. While I was able to get require working, I'm fairly sure the loaded feature index was not being used as expected, which would have significant performance implications. Additionally, I was never able to get that approach working with autoload when autoloading a non-realpath file. It also broke some specs. This takes a more conservative approach. Directly before loading the file, if the file with the same realpath has been required, the loading of the file is skipped. The realpaths are stored as fstrings in a hidden hash. When rebuilding the loaded feature index, the hash of realpaths is also rebuilt. I'm guessing this makes rebuilding process slower, but I don think that is a hot path. In general, modifying loaded features is only done when reloading, and that tends to be in non-production environments. Change test_require_with_loaded_features_pop test to use 30 threads and 300 iterations, instead of 4 threads and 1000 iterations. I saw only sporadic failures with 4/1000, but consistent failures 30/300 threads. These failures were due to the fact that the concurrent deletions from $LOADED_FEATURES in other threads can result in rb_ary_entry returning nil when rebuilding the loaded features index. To avoid concurrency issues when rebuilding the loaded features index, the building of the index itself is left alone, and afterwards, a separate loop is done on a copy of the loaded feature snapshot in order to rebuild the realpaths hash. Fixes [Bug #17885]
2021-06-30 23:50:19 +03:00
}
static VALUE
get_LOADED_FEATURES(ID _x, VALUE *_y)
{
2021-09-16 19:20:10 +03:00
return get_loaded_features(GET_VM());
}
static void
2021-09-16 19:20:10 +03:00
reset_loaded_features_snapshot(rb_vm_t *vm)
{
rb_ary_replace(vm->loaded_features_snapshot, vm->loaded_features);
}
static struct st_table *
2021-09-16 19:20:10 +03:00
get_loaded_features_index_raw(rb_vm_t *vm)
{
2021-09-16 19:20:10 +03:00
return vm->loaded_features_index;
}
static st_table *
2021-09-16 19:20:10 +03:00
get_loading_table(rb_vm_t *vm)
{
2021-09-16 19:20:10 +03:00
return vm->loading_table;
}
static st_data_t
feature_key(const char *str, size_t len)
{
return st_hash(str, len, 0xfea7009e);
}
static bool
is_rbext_path(VALUE feature_path)
{
long len = RSTRING_LEN(feature_path);
long rbext_len = rb_strlen_lit(".rb");
if (len <= rbext_len) return false;
return IS_RBEXT(RSTRING_PTR(feature_path) + len - rbext_len);
}
typedef rb_darray(long) feature_indexes_t;
struct features_index_add_single_args {
rb_vm_t *vm;
VALUE offset;
bool rb;
};
static int
features_index_add_single_callback(st_data_t *key, st_data_t *value, st_data_t raw_args, int existing)
{
struct features_index_add_single_args *args = (struct features_index_add_single_args *)raw_args;
rb_vm_t *vm = args->vm;
VALUE offset = args->offset;
bool rb = args->rb;
if (existing) {
VALUE this_feature_index = *value;
if (FIXNUM_P(this_feature_index)) {
VALUE loaded_features = get_loaded_features(vm);
VALUE this_feature_path = RARRAY_AREF(loaded_features, FIX2LONG(this_feature_index));
feature_indexes_t feature_indexes;
rb_darray_make(&feature_indexes, 2);
int top = (rb && !is_rbext_path(this_feature_path)) ? 1 : 0;
rb_darray_set(feature_indexes, top^0, FIX2LONG(this_feature_index));
rb_darray_set(feature_indexes, top^1, FIX2LONG(offset));
assert(rb_darray_size(feature_indexes) == 2);
// assert feature_indexes does not look like a special const
assert(!SPECIAL_CONST_P((VALUE)feature_indexes));
*value = (st_data_t)feature_indexes;
}
else {
feature_indexes_t feature_indexes = (feature_indexes_t)this_feature_index;
long pos = -1;
if (rb) {
VALUE loaded_features = get_loaded_features(vm);
for (size_t i = 0; i < rb_darray_size(feature_indexes); ++i) {
long idx = rb_darray_get(feature_indexes, i);
VALUE this_feature_path = RARRAY_AREF(loaded_features, idx);
Check_Type(this_feature_path, T_STRING);
if (!is_rbext_path(this_feature_path)) {
pos = i;
break;
}
}
}
rb_darray_append(&feature_indexes, FIX2LONG(offset));
/* darray may realloc which will change the pointer */
*value = (st_data_t)feature_indexes;
if (pos >= 0) {
long *ptr = rb_darray_data_ptr(feature_indexes);
long len = rb_darray_size(feature_indexes);
MEMMOVE(ptr + pos, ptr + pos + 1, long, len - pos - 1);
ptr[pos] = FIX2LONG(offset);
}
}
}
else {
*value = offset;
}
return ST_CONTINUE;
}
static void
2021-09-16 19:20:10 +03:00
features_index_add_single(rb_vm_t *vm, const char* str, size_t len, VALUE offset, bool rb)
{
struct st_table *features_index;
st_data_t short_feature_key;
Check_Type(offset, T_FIXNUM);
short_feature_key = feature_key(str, len);
2021-09-16 19:20:10 +03:00
features_index = get_loaded_features_index_raw(vm);
struct features_index_add_single_args args = {
.vm = vm,
.offset = offset,
.rb = rb,
};
st_update(features_index, short_feature_key, features_index_add_single_callback, (st_data_t)&args);
}
/* Add to the loaded-features index all the required entries for
`feature`, located at `offset` in $LOADED_FEATURES. We add an
index entry at each string `short_feature` for which
feature == "#{prefix}#{short_feature}#{ext}"
where `ext` is empty or matches %r{^\.[^./]*$}, and `prefix` is empty
or ends in '/'. This maintains the invariant that `rb_feature_p()`
relies on for its fast lookup.
*/
static void
2021-09-16 19:20:10 +03:00
features_index_add(rb_vm_t *vm, VALUE feature, VALUE offset)
{
const char *feature_str, *feature_end, *ext, *p;
bool rb = false;
feature_str = StringValuePtr(feature);
feature_end = feature_str + RSTRING_LEN(feature);
for (ext = feature_end; ext > feature_str; ext--)
if (*ext == '.' || *ext == '/')
break;
if (*ext != '.')
ext = NULL;
else
rb = IS_RBEXT(ext);
/* Now `ext` points to the only string matching %r{^\.[^./]*$} that is
at the end of `feature`, or is NULL if there is no such string. */
p = ext ? ext : feature_end;
while (1) {
p--;
while (p >= feature_str && *p != '/')
p--;
if (p < feature_str)
break;
/* Now *p == '/'. We reach this point for every '/' in `feature`. */
2021-09-16 19:20:10 +03:00
features_index_add_single(vm, p + 1, feature_end - p - 1, offset, false);
if (ext) {
2021-09-16 19:20:10 +03:00
features_index_add_single(vm, p + 1, ext - p - 1, offset, rb);
}
}
2021-09-16 19:20:10 +03:00
features_index_add_single(vm, feature_str, feature_end - feature_str, offset, false);
if (ext) {
2021-09-16 19:20:10 +03:00
features_index_add_single(vm, feature_str, ext - feature_str, offset, rb);
}
}
static int
loaded_features_index_clear_i(st_data_t key, st_data_t val, st_data_t arg)
{
VALUE obj = (VALUE)val;
if (!SPECIAL_CONST_P(obj)) {
rb_darray_free((void *)obj);
}
return ST_DELETE;
}
static st_table *
2021-09-16 19:20:10 +03:00
get_loaded_features_index(rb_vm_t *vm)
{
VALUE features;
int i;
if (!rb_ary_shared_with_p(vm->loaded_features_snapshot, vm->loaded_features)) {
/* The sharing was broken; something (other than us in rb_provide_feature())
modified loaded_features. Rebuild the index. */
st_foreach(vm->loaded_features_index, loaded_features_index_clear_i, 0);
Do not load file with same realpath twice when requiring This fixes issues with paths being loaded twice in certain cases when symlinks are used. It took me multiple attempts to get this working. My original attempt tried to convert paths to realpaths before adding them to $LOADED_FEATURES. Unfortunately, this doesn't work well with the loaded feature index, which is based off load paths and not realpaths. While I was able to get require working, I'm fairly sure the loaded feature index was not being used as expected, which would have significant performance implications. Additionally, I was never able to get that approach working with autoload when autoloading a non-realpath file. It also broke some specs. This takes a more conservative approach. Directly before loading the file, if the file with the same realpath has been required, the loading of the file is skipped. The realpaths are stored as fstrings in a hidden hash. When rebuilding the loaded feature index, the hash of realpaths is also rebuilt. I'm guessing this makes rebuilding process slower, but I don think that is a hot path. In general, modifying loaded features is only done when reloading, and that tends to be in non-production environments. Change test_require_with_loaded_features_pop test to use 30 threads and 300 iterations, instead of 4 threads and 1000 iterations. I saw only sporadic failures with 4/1000, but consistent failures 30/300 threads. These failures were due to the fact that the concurrent deletions from $LOADED_FEATURES in other threads can result in rb_ary_entry returning nil when rebuilding the loaded features index. To avoid concurrency issues when rebuilding the loaded features index, the building of the index itself is left alone, and afterwards, a separate loop is done on a copy of the loaded feature snapshot in order to rebuild the realpaths hash. Fixes [Bug #17885]
2021-06-30 23:50:19 +03:00
VALUE realpaths = vm->loaded_features_realpaths;
rb_hash_clear(realpaths);
features = vm->loaded_features;
for (i = 0; i < RARRAY_LEN(features); i++) {
VALUE entry, as_str;
as_str = entry = rb_ary_entry(features, i);
StringValue(as_str);
as_str = rb_fstring(rb_str_freeze(as_str));
if (as_str != entry)
rb_ary_store(features, i, as_str);
2021-09-16 19:20:10 +03:00
features_index_add(vm, as_str, INT2FIX(i));
}
2021-09-16 19:20:10 +03:00
reset_loaded_features_snapshot(vm);
Do not load file with same realpath twice when requiring This fixes issues with paths being loaded twice in certain cases when symlinks are used. It took me multiple attempts to get this working. My original attempt tried to convert paths to realpaths before adding them to $LOADED_FEATURES. Unfortunately, this doesn't work well with the loaded feature index, which is based off load paths and not realpaths. While I was able to get require working, I'm fairly sure the loaded feature index was not being used as expected, which would have significant performance implications. Additionally, I was never able to get that approach working with autoload when autoloading a non-realpath file. It also broke some specs. This takes a more conservative approach. Directly before loading the file, if the file with the same realpath has been required, the loading of the file is skipped. The realpaths are stored as fstrings in a hidden hash. When rebuilding the loaded feature index, the hash of realpaths is also rebuilt. I'm guessing this makes rebuilding process slower, but I don think that is a hot path. In general, modifying loaded features is only done when reloading, and that tends to be in non-production environments. Change test_require_with_loaded_features_pop test to use 30 threads and 300 iterations, instead of 4 threads and 1000 iterations. I saw only sporadic failures with 4/1000, but consistent failures 30/300 threads. These failures were due to the fact that the concurrent deletions from $LOADED_FEATURES in other threads can result in rb_ary_entry returning nil when rebuilding the loaded features index. To avoid concurrency issues when rebuilding the loaded features index, the building of the index itself is left alone, and afterwards, a separate loop is done on a copy of the loaded feature snapshot in order to rebuild the realpaths hash. Fixes [Bug #17885]
2021-06-30 23:50:19 +03:00
features = rb_ary_dup(vm->loaded_features_snapshot);
long j = RARRAY_LEN(features);
for (i = 0; i < j; i++) {
VALUE as_str = rb_ary_entry(features, i);
VALUE realpath = rb_check_realpath(Qnil, as_str, NULL);
if (NIL_P(realpath)) realpath = as_str;
rb_hash_aset(realpaths, rb_fstring(realpath), Qtrue);
}
}
return vm->loaded_features_index;
}
/* This searches `load_path` for a value such that
name == "#{load_path[i]}/#{feature}"
if `feature` is a suffix of `name`, or otherwise
name == "#{load_path[i]}/#{feature}#{ext}"
for an acceptable string `ext`. It returns
`load_path[i].to_str` if found, else 0.
If type is 's', then `ext` is acceptable only if IS_DLEXT(ext);
if 'r', then only if IS_RBEXT(ext); otherwise `ext` may be absent
or have any value matching `%r{^\.[^./]*$}`.
*/
static VALUE
loaded_feature_path(const char *name, long vlen, const char *feature, long len,
int type, VALUE load_path)
{
long i;
long plen;
const char *e;
if (vlen < len+1) return 0;
if (strchr(feature, '.') && !strncmp(name+(vlen-len), feature, len)) {
plen = vlen - len;
}
else {
for (e = name + vlen; name != e && *e != '.' && *e != '/'; --e);
if (*e != '.' ||
e-name < len ||
strncmp(e-len, feature, len))
return 0;
plen = e - name - len;
}
if (plen > 0 && name[plen-1] != '/') {
return 0;
}
if (type == 's' ? !IS_DLEXT(&name[plen+len]) :
type == 'r' ? !IS_RBEXT(&name[plen+len]) :
0) {
return 0;
}
/* Now name == "#{prefix}/#{feature}#{ext}" where ext is acceptable
(possibly empty) and prefix is some string of length plen. */
if (plen > 0) --plen; /* exclude '.' */
for (i = 0; i < RARRAY_LEN(load_path); ++i) {
VALUE p = RARRAY_AREF(load_path, i);
const char *s = StringValuePtr(p);
long n = RSTRING_LEN(p);
if (n != plen) continue;
if (n && strncmp(name, s, n)) continue;
return p;
}
return 0;
}
struct loaded_feature_searching {
const char *name;
long len;
int type;
VALUE load_path;
const char *result;
};
static int
loaded_feature_path_i(st_data_t v, st_data_t b, st_data_t f)
{
const char *s = (const char *)v;
struct loaded_feature_searching *fp = (struct loaded_feature_searching *)f;
VALUE p = loaded_feature_path(s, strlen(s), fp->name, fp->len,
fp->type, fp->load_path);
if (!p) return ST_CONTINUE;
fp->result = s;
return ST_STOP;
}
static int
2021-09-16 19:20:10 +03:00
rb_feature_p(rb_vm_t *vm, const char *feature, const char *ext, int rb, int expanded, const char **fn)
{
VALUE features, this_feature_index = Qnil, v, p, load_path = 0;
const char *f, *e;
long i, len, elen, n;
st_table *loading_tbl, *features_index;
st_data_t data;
st_data_t key;
int type;
if (fn) *fn = 0;
if (ext) {
elen = strlen(ext);
len = strlen(feature) - elen;
type = rb ? 'r' : 's';
}
else {
len = strlen(feature);
elen = 0;
type = 0;
}
2021-09-16 19:20:10 +03:00
features = get_loaded_features(vm);
features_index = get_loaded_features_index(vm);
key = feature_key(feature, strlen(feature));
/* We search `features` for an entry such that either
"#{features[i]}" == "#{load_path[j]}/#{feature}#{e}"
for some j, or
"#{features[i]}" == "#{feature}#{e}"
Here `e` is an "allowed" extension -- either empty or one
of the extensions accepted by IS_RBEXT, IS_SOEXT, or
IS_DLEXT. Further, if `ext && rb` then `IS_RBEXT(e)`,
and if `ext && !rb` then `IS_SOEXT(e) || IS_DLEXT(e)`.
If `expanded`, then only the latter form (without load_path[j])
is accepted. Otherwise either form is accepted, *unless* `ext`
is false and an otherwise-matching entry of the first form is
preceded by an entry of the form
"#{features[i2]}" == "#{load_path[j2]}/#{feature}#{e2}"
where `e2` matches %r{^\.[^./]*$} but is not an allowed extension.
After a "distractor" entry of this form, only entries of the
form "#{feature}#{e}" are accepted.
In `rb_provide_feature()` and `get_loaded_features_index()` we
maintain an invariant that the array `this_feature_index` will
point to every entry in `features` which has the form
"#{prefix}#{feature}#{e}"
where `e` is empty or matches %r{^\.[^./]*$}, and `prefix` is empty
or ends in '/'. This includes both match forms above, as well
as any distractors, so we may ignore all other entries in `features`.
*/
2021-07-13 18:51:53 +03:00
if (st_lookup(features_index, key, &data) && !NIL_P(this_feature_index = (VALUE)data)) {
for (size_t i = 0; ; i++) {
long index;
if (FIXNUM_P(this_feature_index)) {
if (i > 0) break;
index = FIX2LONG(this_feature_index);
}
else {
feature_indexes_t feature_indexes = (feature_indexes_t)this_feature_index;
if (i >= rb_darray_size(feature_indexes)) break;
index = rb_darray_get(feature_indexes, i);
}
v = RARRAY_AREF(features, index);
f = StringValuePtr(v);
if ((n = RSTRING_LEN(v)) < len) continue;
if (strncmp(f, feature, len) != 0) {
if (expanded) continue;
2021-09-16 19:20:10 +03:00
if (!load_path) load_path = get_expanded_load_path(vm);
if (!(p = loaded_feature_path(f, n, feature, len, type, load_path)))
continue;
expanded = 1;
f += RSTRING_LEN(p) + 1;
}
if (!*(e = f + len)) {
if (ext) continue;
return 'u';
}
if (*e != '.') continue;
if ((!rb || !ext) && (IS_SOEXT(e) || IS_DLEXT(e))) {
return 's';
}
if ((rb || !ext) && (IS_RBEXT(e))) {
return 'r';
}
}
}
2021-09-16 19:20:10 +03:00
loading_tbl = get_loading_table(vm);
f = 0;
if (!expanded) {
struct loaded_feature_searching fs;
fs.name = feature;
fs.len = len;
fs.type = type;
2021-09-16 19:20:10 +03:00
fs.load_path = load_path ? load_path : get_expanded_load_path(vm);
fs.result = 0;
st_foreach(loading_tbl, loaded_feature_path_i, (st_data_t)&fs);
if ((f = fs.result) != 0) {
if (fn) *fn = f;
goto loading;
}
}
if (st_get_key(loading_tbl, (st_data_t)feature, &data)) {
if (fn) *fn = (const char*)data;
goto loading;
}
else {
VALUE bufstr;
char *buf;
static const char so_ext[][4] = {
".so", ".o",
};
2022-07-21 19:23:58 +03:00
if (ext && *ext) return 0;
bufstr = rb_str_tmp_new(len + DLEXT_MAXLEN);
buf = RSTRING_PTR(bufstr);
MEMCPY(buf, feature, char, len);
for (i = 0; (e = loadable_ext[i]) != 0; i++) {
strlcpy(buf + len, e, DLEXT_MAXLEN + 1);
if (st_get_key(loading_tbl, (st_data_t)buf, &data)) {
rb_str_resize(bufstr, 0);
if (fn) *fn = (const char*)data;
return i ? 's' : 'r';
}
}
for (i = 0; i < numberof(so_ext); i++) {
strlcpy(buf + len, so_ext[i], DLEXT_MAXLEN + 1);
if (st_get_key(loading_tbl, (st_data_t)buf, &data)) {
rb_str_resize(bufstr, 0);
if (fn) *fn = (const char*)data;
return 's';
}
}
rb_str_resize(bufstr, 0);
}
return 0;
loading:
if (!ext) return 'u';
return !IS_RBEXT(ext) ? 's' : 'r';
}
int
rb_provided(const char *feature)
{
return rb_feature_provided(feature, 0);
}
2021-09-16 19:20:10 +03:00
static int
feature_provided(rb_vm_t *vm, const char *feature, const char **loading)
{
const char *ext = strrchr(feature, '.');
VALUE fullpath = 0;
if (*feature == '.' &&
(feature[1] == '/' || strncmp(feature+1, "./", 2) == 0)) {
Fix compatibility of cached expanded load path * file.c (rb_get_path_check_to_string): extract from rb_get_path_check(). We change the spec not to call to_path of String object. * file.c (rb_get_path_check_convert): extract from rb_get_path_check(). * file.c (rb_get_path_check): follow the above change. * file.c (rb_file_expand_path_fast): remove check_expand_path_args(). Instead we call it in load.c. * file.c (rb_find_file_ext_safe): use rb_get_expanded_load_path() to reduce expand cost. * file.c (rb_find_file_safe): ditto. * internal.h (rb_get_expanded_load_path): add a declaration. * internal.h (rb_get_path_check_to_string, rb_get_path_check_convert): add declarations. * load.c (rb_construct_expanded_load_path): fix for compatibility. Same checks in rb_get_path_check() are added. We don't replace $LOAD_PATH and ensure that String object of $LOAD_PATH are frozen. We don't freeze non String object and expand it every times. We add arguments for expanding load path partially and checking if load path have relative paths or non String objects. * load.c (load_path_getcwd): get current working directory for checking if it's changed when getting load path. * load.c (rb_get_expanded_load_path): fix for rebuilding cache properly. We check if current working directory is changed and rebuild expanded load path cache. We expand paths which start with ~ (User HOME) and non String objects every times for compatibility. We make this accessible from other source files. * load.c (rb_feature_provided): call rb_get_path() since we changed rb_file_expand_path_fast() not to call it. * load.c (Init_load): initialize vm->load_path_check_cache. * vm.c (rb_vm_mark): mark vm->load_path_check_cache for GC. * vm_core.h (rb_vm_struct): add vm->load_path_check_cache to store data to check load path cache validity. * test/ruby/test_require.rb (TestRequire): add tests for require compatibility related to cached expanded load path. [ruby-core:47970] [Bug #7158] git-svn-id: svn+ssh://ci.ruby-lang.org/ruby/trunk@37482 b2dd03c8-39d4-4d8f-98ff-823fe69b080e
2012-11-05 19:27:08 +04:00
fullpath = rb_file_expand_path_fast(rb_get_path(rb_str_new2(feature)), Qnil);
feature = RSTRING_PTR(fullpath);
}
if (ext && !strchr(ext, '/')) {
if (IS_RBEXT(ext)) {
2021-09-16 19:20:10 +03:00
if (rb_feature_p(vm, feature, ext, TRUE, FALSE, loading)) return TRUE;
return FALSE;
}
else if (IS_SOEXT(ext) || IS_DLEXT(ext)) {
2021-09-16 19:20:10 +03:00
if (rb_feature_p(vm, feature, ext, FALSE, FALSE, loading)) return TRUE;
return FALSE;
}
}
2021-09-16 19:20:10 +03:00
if (rb_feature_p(vm, feature, 0, TRUE, FALSE, loading))
return TRUE;
RB_GC_GUARD(fullpath);
return FALSE;
}
2021-09-16 19:20:10 +03:00
int
rb_feature_provided(const char *feature, const char **loading)
{
return feature_provided(GET_VM(), feature, loading);
}
static void
2021-09-16 19:20:10 +03:00
rb_provide_feature(rb_vm_t *vm, VALUE feature)
{
VALUE features;
2021-09-16 19:20:10 +03:00
features = get_loaded_features(vm);
if (OBJ_FROZEN(features)) {
rb_raise(rb_eRuntimeError,
"$LOADED_FEATURES is frozen; cannot append feature");
}
rb_str_freeze(feature);
2021-09-16 19:20:10 +03:00
get_loaded_features_index(vm);
// If loaded_features and loaded_features_snapshot share the same backing
// array, pushing into it would cause the whole array to be copied.
// To avoid this we first clear loaded_features_snapshot.
rb_ary_clear(vm->loaded_features_snapshot);
rb_ary_push(features, rb_fstring(feature));
2021-09-16 19:20:10 +03:00
features_index_add(vm, feature, INT2FIX(RARRAY_LEN(features)-1));
reset_loaded_features_snapshot(vm);
}
void
rb_provide(const char *feature)
{
2021-09-16 19:20:10 +03:00
rb_provide_feature(GET_VM(), rb_fstring_cstr(feature));
}
NORETURN(static void load_failed(VALUE));
* this commit is a result of refactoring. only renaming functions, moving definitions place, add/remove prototypes, deleting unused variables and removing yarv.h. This commit doesn't change any behavior of ruby/vm. * yarv.h, common.mk: remove yarv.h (contents are moved to yarvcore.h). * error.c, eval_intern.h: include yarvcore.h instead yarv.h * rename some functions: * debug.[ch]: debug_*() -> ruby_debug_*() * iseq.c: iseq_*() -> rb_iseq_*(), ruby_iseq_disasm() * iseq.c: node_name() -> ruby_node_name() * vm.c: yarv_check_redefinition_opt_method() -> rb_vm_check_redefinition_opt_method() * some refactoring with checking -Wall. * array.c: remove rb_ary_ptr() (unused) and remove unused local variables. * object.c: add a prototype of rb_mod_module_exec(). * eval_intern.h (ruby_cref): set it inline. * eval_load.c (rb_load), yarvcore.c: yarv_load() -> rb_load_internal(). * parse.y: add a prototype of rb_parse_in_eval() (in eval.c). * process.c: add a prototype of rb_thread_stop_timer_thread() (in thread.c). * thread.c: remove raw_gets() function (unused) and fix some format mismatch (format mismatchs have remained yet. this is todo). * thread.c (rb_thread_wait_fd_rw): fix typo on label name. * thread_pthread.ci: comment out codes with USE_THREAD_CACHE. * vm.c (rb_svar, rb_backref_get, rb_backref_get, rb_lastline_get, rb_lastline_set) : moved from yarvcore.c. * vm.c (yarv_init_redefined_flag): add a prototype and rename yarv_opt_method_table to vm_opt_method_table. * vm.c (rb_thread_eval): moved from yarvcore.c. * yarvcore.c: remove unused global variables and fix to use nsdr(). git-svn-id: svn+ssh://ci.ruby-lang.org/ruby/trunk@11652 b2dd03c8-39d4-4d8f-98ff-823fe69b080e
2007-02-07 04:25:05 +03:00
static inline void
2019-08-08 16:17:22 +03:00
load_iseq_eval(rb_execution_context_t *ec, VALUE fname)
{
const rb_iseq_t *iseq = rb_iseq_load_iseq(fname);
if (!iseq) {
rb_ast_t *ast;
VALUE parser = rb_parser_new();
rb_parser_set_context(parser, NULL, FALSE);
ast = (rb_ast_t *)rb_parser_load_file(parser, fname);
iseq = rb_iseq_new_top(&ast->body, rb_fstring_lit("<top (required)>"),
fname, rb_realpath_internal(Qnil, fname, 1), NULL);
rb_ast_dispose(ast);
}
rb_exec_event_hook_script_compiled(ec, iseq, Qnil);
rb_iseq_eval(iseq);
}
static inline enum ruby_tag_type
load_wrapping(rb_execution_context_t *ec, VALUE fname, VALUE load_wrapper)
{
enum ruby_tag_type state;
rb_thread_t *th = rb_ec_thread_ptr(ec);
volatile VALUE wrapper = th->top_wrapper;
volatile VALUE self = th->top_self;
#if !defined __GNUC__
rb_thread_t *volatile th0 = th;
#endif
ec->errinfo = Qnil; /* ensure */
/* load in module as toplevel */
th->top_self = rb_obj_clone(rb_vm_top_self());
th->top_wrapper = load_wrapper;
rb_extend_object(th->top_self, th->top_wrapper);
EC_PUSH_TAG(ec);
state = EC_EXEC_TAG();
if (state == TAG_NONE) {
2019-08-08 16:17:22 +03:00
load_iseq_eval(ec, fname);
}
EC_POP_TAG();
#if !defined __GNUC__
th = th0;
fname = RB_GC_GUARD(fname);
#endif
th->top_self = self;
th->top_wrapper = wrapper;
2019-08-08 16:17:22 +03:00
return state;
}
2019-08-08 16:17:22 +03:00
static inline void
raise_load_if_failed(rb_execution_context_t *ec, enum ruby_tag_type state)
{
if (state) {
rb_vm_jump_tag_but_local_jump(state);
}
if (!NIL_P(ec->errinfo)) {
rb_exc_raise(ec->errinfo);
}
}
static void
rb_load_internal(VALUE fname, VALUE wrap)
{
rb_execution_context_t *ec = GET_EC();
enum ruby_tag_type state = TAG_NONE;
if (RTEST(wrap)) {
if (!RB_TYPE_P(wrap, T_MODULE)) {
wrap = rb_module_new();
}
state = load_wrapping(ec, fname, wrap);
}
else {
load_iseq_eval(ec, fname);
}
raise_load_if_failed(ec, state);
}
void
rb_load(VALUE fname, int wrap)
{
VALUE tmp = rb_find_file(FilePathValue(fname));
if (!tmp) load_failed(fname);
rb_load_internal(tmp, RBOOL(wrap));
}
void
rb_load_protect(VALUE fname, int wrap, int *pstate)
{
enum ruby_tag_type state;
EC_PUSH_TAG(GET_EC());
if ((state = EC_EXEC_TAG()) == TAG_NONE) {
rb_load(fname, wrap);
}
EC_POP_TAG();
if (state != TAG_NONE) *pstate = state;
}
/*
* call-seq:
* load(filename, wrap=false) -> true
*
* Loads and executes the Ruby program in the file _filename_.
*
* If the filename is an absolute path (e.g. starts with '/'), the file
* will be loaded directly using the absolute path.
*
* If the filename is an explicit relative path (e.g. starts with './' or
* '../'), the file will be loaded using the relative path from the current
* directory.
*
* Otherwise, the file will be searched for in the library
* directories listed in <code>$LOAD_PATH</code> (<code>$:</code>).
* If the file is found in a directory, it will attempt to load the file
* relative to that directory. If the file is not found in any of the
* directories in <code>$LOAD_PATH</code>, the file will be loaded using
* the relative path from the current directory.
*
* If the file doesn't exist when there is an attempt to load it, a
* LoadError will be raised.
*
* If the optional _wrap_ parameter is +true+, the loaded script will
* be executed under an anonymous module, protecting the calling
* program's global namespace. If the optional _wrap_ parameter is a
* module, the loaded script will be executed under the given module.
* In no circumstance will any local variables in the loaded file be
* propagated to the loading environment.
*/
static VALUE
rb_f_load(int argc, VALUE *argv, VALUE _)
{
VALUE fname, wrap, path, orig_fname;
rb_scan_args(argc, argv, "11", &fname, &wrap);
* probes.d: add DTrace probe declarations. [ruby-core:27448] * array.c (empty_ary_alloc, ary_new): added array create DTrace probe. * compile.c (rb_insns_name): allowing DTrace probes to access instruction sequence name. * Makefile.in: translate probes.d file to appropriate header file. * common.mk: declare dependencies on the DTrace header. * configure.in: add a test for existence of DTrace. * eval.c (setup_exception): add a probe for when an exception is raised. * gc.c: Add DTrace probes for mark begin and end, and sweep begin and end. * hash.c (empty_hash_alloc): Add a probe for hash allocation. * insns.def: Add probes for function entry and return. * internal.h: function declaration for compile.c change. * load.c (rb_f_load): add probes for `load` entry and exit, require entry and exit, and wrapping search_required for load path search. * object.c (rb_obj_alloc): added a probe for general object creation. * parse.y (yycompile0): added a probe around parse and compile phase. * string.c (empty_str_alloc, str_new): DTrace probes for string allocation. * test/dtrace/*: tests for DTrace probes. * vm.c (vm_invoke_proc): add probes for function return on exception raise, hash create, and instruction sequence execution. * vm_core.h: add probe declarations for function entry and exit. * vm_dump.c: add probes header file. * vm_eval.c (vm_call0_cfunc, vm_call0_cfunc_with_frame): add probe on function entry and return. * vm_exec.c: expose instruction number to instruction name function. * vm_insnshelper.c: add function entry and exit probes for cfunc methods. * vm_insnhelper.h: vm usage information is always collected, so uncomment the functions. 12 19:14:50 2012 Akinori MUSHA <knu@iDaemons.org> * configure.in (isinf, isnan): isinf() and isnan() are macros on DragonFly which cannot be found by AC_REPLACE_FUNCS(). This workaround enforces the fact that they exist on DragonFly. 12 15:59:38 2012 Shugo Maeda <shugo@ruby-lang.org> * vm_core.h (rb_call_info_t::refinements), compile.c (new_callinfo), vm_insnhelper.c (vm_search_method): revert r37616 because it's too slow. [ruby-dev:46477] * test/ruby/test_refinement.rb (test_inline_method_cache): skip the test until the bug is fixed efficiently. git-svn-id: svn+ssh://ci.ruby-lang.org/ruby/trunk@37631 b2dd03c8-39d4-4d8f-98ff-823fe69b080e
2012-11-13 01:52:12 +04:00
orig_fname = rb_get_path_check_to_string(fname);
fname = rb_str_encode_ospath(orig_fname);
RUBY_DTRACE_HOOK(LOAD_ENTRY, RSTRING_PTR(orig_fname));
path = rb_find_file(fname);
if (!path) {
if (!rb_file_load_ok(RSTRING_PTR(fname)))
load_failed(orig_fname);
path = fname;
}
rb_load_internal(path, wrap);
* probes.d: add DTrace probe declarations. [ruby-core:27448] * array.c (empty_ary_alloc, ary_new): added array create DTrace probe. * compile.c (rb_insns_name): allowing DTrace probes to access instruction sequence name. * Makefile.in: translate probes.d file to appropriate header file. * common.mk: declare dependencies on the DTrace header. * configure.in: add a test for existence of DTrace. * eval.c (setup_exception): add a probe for when an exception is raised. * gc.c: Add DTrace probes for mark begin and end, and sweep begin and end. * hash.c (empty_hash_alloc): Add a probe for hash allocation. * insns.def: Add probes for function entry and return. * internal.h: function declaration for compile.c change. * load.c (rb_f_load): add probes for `load` entry and exit, require entry and exit, and wrapping search_required for load path search. * object.c (rb_obj_alloc): added a probe for general object creation. * parse.y (yycompile0): added a probe around parse and compile phase. * string.c (empty_str_alloc, str_new): DTrace probes for string allocation. * test/dtrace/*: tests for DTrace probes. * vm.c (vm_invoke_proc): add probes for function return on exception raise, hash create, and instruction sequence execution. * vm_core.h: add probe declarations for function entry and exit. * vm_dump.c: add probes header file. * vm_eval.c (vm_call0_cfunc, vm_call0_cfunc_with_frame): add probe on function entry and return. * vm_exec.c: expose instruction number to instruction name function. * vm_insnshelper.c: add function entry and exit probes for cfunc methods. * vm_insnhelper.h: vm usage information is always collected, so uncomment the functions. 12 19:14:50 2012 Akinori MUSHA <knu@iDaemons.org> * configure.in (isinf, isnan): isinf() and isnan() are macros on DragonFly which cannot be found by AC_REPLACE_FUNCS(). This workaround enforces the fact that they exist on DragonFly. 12 15:59:38 2012 Shugo Maeda <shugo@ruby-lang.org> * vm_core.h (rb_call_info_t::refinements), compile.c (new_callinfo), vm_insnhelper.c (vm_search_method): revert r37616 because it's too slow. [ruby-dev:46477] * test/ruby/test_refinement.rb (test_inline_method_cache): skip the test until the bug is fixed efficiently. git-svn-id: svn+ssh://ci.ruby-lang.org/ruby/trunk@37631 b2dd03c8-39d4-4d8f-98ff-823fe69b080e
2012-11-13 01:52:12 +04:00
RUBY_DTRACE_HOOK(LOAD_RETURN, RSTRING_PTR(orig_fname));
* probes.d: add DTrace probe declarations. [ruby-core:27448] * array.c (empty_ary_alloc, ary_new): added array create DTrace probe. * compile.c (rb_insns_name): allowing DTrace probes to access instruction sequence name. * Makefile.in: translate probes.d file to appropriate header file. * common.mk: declare dependencies on the DTrace header. * configure.in: add a test for existence of DTrace. * eval.c (setup_exception): add a probe for when an exception is raised. * gc.c: Add DTrace probes for mark begin and end, and sweep begin and end. * hash.c (empty_hash_alloc): Add a probe for hash allocation. * insns.def: Add probes for function entry and return. * internal.h: function declaration for compile.c change. * load.c (rb_f_load): add probes for `load` entry and exit, require entry and exit, and wrapping search_required for load path search. * object.c (rb_obj_alloc): added a probe for general object creation. * parse.y (yycompile0): added a probe around parse and compile phase. * string.c (empty_str_alloc, str_new): DTrace probes for string allocation. * test/dtrace/*: tests for DTrace probes. * vm.c (vm_invoke_proc): add probes for function return on exception raise, hash create, and instruction sequence execution. * vm_core.h: add probe declarations for function entry and exit. * vm_dump.c: add probes header file. * vm_eval.c (vm_call0_cfunc, vm_call0_cfunc_with_frame): add probe on function entry and return. * vm_exec.c: expose instruction number to instruction name function. * vm_insnshelper.c: add function entry and exit probes for cfunc methods. * vm_insnhelper.h: vm usage information is always collected, so uncomment the functions. 12 19:14:50 2012 Akinori MUSHA <knu@iDaemons.org> * configure.in (isinf, isnan): isinf() and isnan() are macros on DragonFly which cannot be found by AC_REPLACE_FUNCS(). This workaround enforces the fact that they exist on DragonFly. 12 15:59:38 2012 Shugo Maeda <shugo@ruby-lang.org> * vm_core.h (rb_call_info_t::refinements), compile.c (new_callinfo), vm_insnhelper.c (vm_search_method): revert r37616 because it's too slow. [ruby-dev:46477] * test/ruby/test_refinement.rb (test_inline_method_cache): skip the test until the bug is fixed efficiently. git-svn-id: svn+ssh://ci.ruby-lang.org/ruby/trunk@37631 b2dd03c8-39d4-4d8f-98ff-823fe69b080e
2012-11-13 01:52:12 +04:00
return Qtrue;
}
static char *
2021-09-16 19:20:10 +03:00
load_lock(rb_vm_t *vm, const char *ftptr, bool warn)
{
st_data_t data;
2021-09-16 19:20:10 +03:00
st_table *loading_tbl = get_loading_table(vm);
if (!st_lookup(loading_tbl, (st_data_t)ftptr, &data)) {
/* partial state */
ftptr = ruby_strdup(ftptr);
data = (st_data_t)rb_thread_shield_new();
st_insert(loading_tbl, (st_data_t)ftptr, data);
return (char *)ftptr;
}
else if (imemo_type_p(data, imemo_memo)) {
struct MEMO *memo = MEMO_CAST(data);
void (*init)(void) = memo->u3.func;
data = (st_data_t)rb_thread_shield_new();
st_insert(loading_tbl, (st_data_t)ftptr, data);
(*init)();
return (char *)"";
}
if (warn) {
VALUE warning = rb_warning_string("loading in progress, circular require considered harmful - %s", ftptr);
rb_backtrace_each(rb_str_append, warning);
rb_warning("%"PRIsVALUE, warning);
}
switch (rb_thread_shield_wait((VALUE)data)) {
case Qfalse:
case Qnil:
return 0;
}
return (char *)ftptr;
}
static int
release_thread_shield(st_data_t *key, st_data_t *value, st_data_t done, int existing)
{
VALUE thread_shield = (VALUE)*value;
if (!existing) return ST_STOP;
if (done) {
rb_thread_shield_destroy(thread_shield);
/* Delete the entry even if there are waiting threads, because they
* won't load the file and won't delete the entry. */
}
else if (rb_thread_shield_release(thread_shield)) {
/* still in-use */
return ST_CONTINUE;
}
xfree((char *)*key);
return ST_DELETE;
}
static void
2021-09-16 19:20:10 +03:00
load_unlock(rb_vm_t *vm, const char *ftptr, int done)
{
if (ftptr) {
st_data_t key = (st_data_t)ftptr;
2021-09-16 19:20:10 +03:00
st_table *loading_tbl = get_loading_table(vm);
st_update(loading_tbl, key, release_thread_shield, done);
}
}
/*
* call-seq:
* require(name) -> true or false
*
* Loads the given +name+, returning +true+ if successful and +false+ if the
* feature is already loaded.
*
* If the filename neither resolves to an absolute path nor starts with
* './' or '../', the file will be searched for in the library
* directories listed in <code>$LOAD_PATH</code> (<code>$:</code>).
* If the filename starts with './' or '../', resolution is based on Dir.pwd.
*
* If the filename has the extension ".rb", it is loaded as a source file; if
* the extension is ".so", ".o", or ".dll", or the default shared library
* extension on the current platform, Ruby loads the shared library as a
* Ruby extension. Otherwise, Ruby tries adding ".rb", ".so", and so on
* to the name until found. If the file named cannot be found, a LoadError
* will be raised.
*
* For Ruby extensions the filename given may use any shared library
* extension. For example, on Linux the socket extension is "socket.so" and
* <code>require 'socket.dll'</code> will load the socket extension.
*
* The absolute path of the loaded file is added to
* <code>$LOADED_FEATURES</code> (<code>$"</code>). A file will not be
* loaded again if its path already appears in <code>$"</code>. For example,
* <code>require 'a'; require './a'</code> will not load <code>a.rb</code>
* again.
*
* require "my-library.rb"
* require "db-driver"
*
* Any constants or globals within the loaded source file will be available
* in the calling program's global namespace. However, local variables will
* not be propagated to the loading environment.
*
*/
VALUE
rb_f_require(VALUE obj, VALUE fname)
{
return rb_require_string(fname);
}
/*
* call-seq:
* require_relative(string) -> true or false
*
* Ruby tries to load the library named _string_ relative to the directory
* containing the requiring file. If the file does not exist a LoadError is
* raised. Returns +true+ if the file was loaded and +false+ if the file was
* already loaded before.
*/
VALUE
rb_f_require_relative(VALUE obj, VALUE fname)
{
VALUE base = rb_current_realfilepath();
if (NIL_P(base)) {
rb_loaderror("cannot infer basepath");
}
base = rb_file_dirname(base);
return rb_require_string(rb_file_absolute_path(fname, base));
}
2021-09-16 19:20:10 +03:00
typedef int (*feature_func)(rb_vm_t *vm, const char *feature, const char *ext, int rb, int expanded, const char **fn);
static int
2021-09-16 19:20:10 +03:00
search_required(rb_vm_t *vm, VALUE fname, volatile VALUE *path, feature_func rb_feature_p)
{
VALUE tmp;
char *ext, *ftptr;
int type, ft = 0;
const char *loading;
*path = 0;
ext = strrchr(ftptr = RSTRING_PTR(fname), '.');
if (ext && !strchr(ext, '/')) {
if (IS_RBEXT(ext)) {
2021-09-16 19:20:10 +03:00
if (rb_feature_p(vm, ftptr, ext, TRUE, FALSE, &loading)) {
if (loading) *path = rb_filesystem_str_new_cstr(loading);
return 'r';
}
if ((tmp = rb_find_file(fname)) != 0) {
ext = strrchr(ftptr = RSTRING_PTR(tmp), '.');
2021-09-16 19:20:10 +03:00
if (!rb_feature_p(vm, ftptr, ext, TRUE, TRUE, &loading) || loading)
*path = tmp;
return 'r';
}
return 0;
}
else if (IS_SOEXT(ext)) {
2021-09-16 19:20:10 +03:00
if (rb_feature_p(vm, ftptr, ext, FALSE, FALSE, &loading)) {
if (loading) *path = rb_filesystem_str_new_cstr(loading);
return 's';
}
tmp = rb_str_subseq(fname, 0, ext - RSTRING_PTR(fname));
rb_str_cat2(tmp, DLEXT);
OBJ_FREEZE(tmp);
if ((tmp = rb_find_file(tmp)) != 0) {
ext = strrchr(ftptr = RSTRING_PTR(tmp), '.');
2021-09-16 19:20:10 +03:00
if (!rb_feature_p(vm, ftptr, ext, FALSE, TRUE, &loading) || loading)
*path = tmp;
return 's';
}
}
else if (IS_DLEXT(ext)) {
2021-09-16 19:20:10 +03:00
if (rb_feature_p(vm, ftptr, ext, FALSE, FALSE, &loading)) {
if (loading) *path = rb_filesystem_str_new_cstr(loading);
return 's';
}
if ((tmp = rb_find_file(fname)) != 0) {
ext = strrchr(ftptr = RSTRING_PTR(tmp), '.');
2021-09-16 19:20:10 +03:00
if (!rb_feature_p(vm, ftptr, ext, FALSE, TRUE, &loading) || loading)
*path = tmp;
return 's';
}
}
}
2021-09-16 19:20:10 +03:00
else if ((ft = rb_feature_p(vm, ftptr, 0, FALSE, FALSE, &loading)) == 'r') {
if (loading) *path = rb_filesystem_str_new_cstr(loading);
return 'r';
}
tmp = fname;
type = rb_find_file_ext(&tmp, ft == 's' ? ruby_ext : loadable_ext);
switch (type) {
case 0:
if (ft)
goto statically_linked;
ftptr = RSTRING_PTR(tmp);
2021-09-16 19:20:10 +03:00
return rb_feature_p(vm, ftptr, 0, FALSE, TRUE, 0);
default:
if (ft) {
goto statically_linked;
}
2019-07-14 11:32:49 +03:00
/* fall through */
case 1:
ext = strrchr(ftptr = RSTRING_PTR(tmp), '.');
2021-09-16 19:20:10 +03:00
if (rb_feature_p(vm, ftptr, ext, !--type, TRUE, &loading) && !loading)
break;
*path = tmp;
}
return type ? 's' : 'r';
statically_linked:
if (loading) *path = rb_filesystem_str_new_cstr(loading);
return ft;
}
static void
load_failed(VALUE fname)
{
rb_load_fail(fname, "cannot load such file");
}
static VALUE
load_ext(VALUE path)
{
rb_scope_visibility_set(METHOD_VISI_PUBLIC);
return (VALUE)dln_load(RSTRING_PTR(path));
}
static int
2021-09-16 19:20:10 +03:00
no_feature_p(rb_vm_t *vm, const char *feature, const char *ext, int rb, int expanded, const char **fn)
{
return 0;
}
// Documented in doc/globals.rdoc
VALUE
rb_resolve_feature_path(VALUE klass, VALUE fname)
{
VALUE path;
int found;
VALUE sym;
fname = rb_get_path(fname);
path = rb_str_encode_ospath(fname);
2021-09-16 19:20:10 +03:00
found = search_required(GET_VM(), path, &path, no_feature_p);
switch (found) {
case 'r':
sym = ID2SYM(rb_intern("rb"));
break;
case 's':
sym = ID2SYM(rb_intern("so"));
break;
default:
return Qnil;
}
return rb_ary_new_from_args(2, sym, path);
}
static void
ext_config_push(rb_thread_t *th, struct rb_ext_config *prev)
{
*prev = th->ext_config;
th->ext_config = (struct rb_ext_config){0};
}
static void
ext_config_pop(rb_thread_t *th, struct rb_ext_config *prev)
{
th->ext_config = *prev;
}
void
rb_ext_ractor_safe(bool flag)
{
GET_THREAD()->ext_config.ractor_safe = flag;
}
/*
* returns
* 0: if already loaded (false)
* 1: successfully loaded (true)
* <0: not found (LoadError)
* >1: exception
*/
2019-08-08 16:17:22 +03:00
static int
require_internal(rb_execution_context_t *ec, VALUE fname, int exception, bool warn)
{
volatile int result = -1;
2019-08-08 18:06:08 +03:00
rb_thread_t *th = rb_ec_thread_ptr(ec);
2021-06-14 07:52:26 +03:00
volatile const struct {
VALUE wrapper, self, errinfo;
} saved = {
th->top_wrapper, th->top_self, ec->errinfo,
};
enum ruby_tag_type state;
char *volatile ftptr = 0;
VALUE path;
2021-06-14 07:49:10 +03:00
volatile VALUE saved_path;
volatile VALUE realpath = 0;
2021-09-16 19:20:10 +03:00
VALUE realpaths = get_loaded_features_realpaths(th->vm);
volatile bool reset_ext_config = false;
struct rb_ext_config prev_ext_config;
fname = rb_get_path(fname);
path = rb_str_encode_ospath(fname);
RUBY_DTRACE_HOOK(REQUIRE_ENTRY, RSTRING_PTR(fname));
2021-06-14 07:49:10 +03:00
saved_path = path;
EC_PUSH_TAG(ec);
2019-08-08 16:17:22 +03:00
ec->errinfo = Qnil; /* ensure */
2019-08-08 18:06:08 +03:00
th->top_wrapper = 0;
if ((state = EC_EXEC_TAG()) == TAG_NONE) {
long handle;
int found;
RUBY_DTRACE_HOOK(FIND_REQUIRE_ENTRY, RSTRING_PTR(fname));
2021-09-16 19:20:10 +03:00
found = search_required(th->vm, path, &saved_path, rb_feature_p);
RUBY_DTRACE_HOOK(FIND_REQUIRE_RETURN, RSTRING_PTR(fname));
2021-06-14 07:49:10 +03:00
path = saved_path;
if (found) {
2021-09-16 19:20:10 +03:00
if (!path || !(ftptr = load_lock(th->vm, RSTRING_PTR(path), warn))) {
result = 0;
}
else if (!*ftptr) {
result = TAG_RETURN;
}
Do not load file with same realpath twice when requiring This fixes issues with paths being loaded twice in certain cases when symlinks are used. It took me multiple attempts to get this working. My original attempt tried to convert paths to realpaths before adding them to $LOADED_FEATURES. Unfortunately, this doesn't work well with the loaded feature index, which is based off load paths and not realpaths. While I was able to get require working, I'm fairly sure the loaded feature index was not being used as expected, which would have significant performance implications. Additionally, I was never able to get that approach working with autoload when autoloading a non-realpath file. It also broke some specs. This takes a more conservative approach. Directly before loading the file, if the file with the same realpath has been required, the loading of the file is skipped. The realpaths are stored as fstrings in a hidden hash. When rebuilding the loaded feature index, the hash of realpaths is also rebuilt. I'm guessing this makes rebuilding process slower, but I don think that is a hot path. In general, modifying loaded features is only done when reloading, and that tends to be in non-production environments. Change test_require_with_loaded_features_pop test to use 30 threads and 300 iterations, instead of 4 threads and 1000 iterations. I saw only sporadic failures with 4/1000, but consistent failures 30/300 threads. These failures were due to the fact that the concurrent deletions from $LOADED_FEATURES in other threads can result in rb_ary_entry returning nil when rebuilding the loaded features index. To avoid concurrency issues when rebuilding the loaded features index, the building of the index itself is left alone, and afterwards, a separate loop is done on a copy of the loaded feature snapshot in order to rebuild the realpaths hash. Fixes [Bug #17885]
2021-06-30 23:50:19 +03:00
else if (RTEST(rb_hash_aref(realpaths,
realpath = rb_realpath_internal(Qnil, path, 1)))) {
Do not load file with same realpath twice when requiring This fixes issues with paths being loaded twice in certain cases when symlinks are used. It took me multiple attempts to get this working. My original attempt tried to convert paths to realpaths before adding them to $LOADED_FEATURES. Unfortunately, this doesn't work well with the loaded feature index, which is based off load paths and not realpaths. While I was able to get require working, I'm fairly sure the loaded feature index was not being used as expected, which would have significant performance implications. Additionally, I was never able to get that approach working with autoload when autoloading a non-realpath file. It also broke some specs. This takes a more conservative approach. Directly before loading the file, if the file with the same realpath has been required, the loading of the file is skipped. The realpaths are stored as fstrings in a hidden hash. When rebuilding the loaded feature index, the hash of realpaths is also rebuilt. I'm guessing this makes rebuilding process slower, but I don think that is a hot path. In general, modifying loaded features is only done when reloading, and that tends to be in non-production environments. Change test_require_with_loaded_features_pop test to use 30 threads and 300 iterations, instead of 4 threads and 1000 iterations. I saw only sporadic failures with 4/1000, but consistent failures 30/300 threads. These failures were due to the fact that the concurrent deletions from $LOADED_FEATURES in other threads can result in rb_ary_entry returning nil when rebuilding the loaded features index. To avoid concurrency issues when rebuilding the loaded features index, the building of the index itself is left alone, and afterwards, a separate loop is done on a copy of the loaded feature snapshot in order to rebuild the realpaths hash. Fixes [Bug #17885]
2021-06-30 23:50:19 +03:00
result = 0;
}
else {
switch (found) {
case 'r':
2019-08-08 16:17:22 +03:00
load_iseq_eval(ec, path);
break;
case 's':
reset_ext_config = true;
ext_config_push(th, &prev_ext_config);
handle = (long)rb_vm_call_cfunc(rb_vm_top_self(), load_ext,
path, VM_BLOCK_HANDLER_NONE, path);
rb_ary_push(ruby_dln_librefs, LONG2NUM(handle));
break;
}
result = TAG_RETURN;
}
}
}
EC_POP_TAG();
rb_thread_t *th2 = rb_ec_thread_ptr(ec);
2021-06-14 07:52:26 +03:00
th2->top_self = saved.self;
th2->top_wrapper = saved.wrapper;
if (reset_ext_config) ext_config_pop(th2, &prev_ext_config);
2021-06-14 07:49:10 +03:00
path = saved_path;
2021-09-16 19:20:10 +03:00
if (ftptr) load_unlock(th2->vm, RSTRING_PTR(path), !state);
if (state) {
if (state == TAG_FATAL || state == TAG_THROW) {
EC_JUMP_TAG(ec, state);
}
else if (exception) {
2019-08-08 16:17:22 +03:00
/* usually state == TAG_RAISE only, except for
* rb_iseq_load_iseq in load_iseq_eval case */
VALUE exc = rb_vm_make_jump_tag_but_local_jump(state, Qundef);
if (!NIL_P(exc)) ec->errinfo = exc;
return TAG_RAISE;
}
else if (state == TAG_RETURN) {
return TAG_RAISE;
}
RB_GC_GUARD(fname);
/* never TAG_RETURN */
return state;
}
2019-08-08 16:17:22 +03:00
if (!NIL_P(ec->errinfo)) {
if (!exception) return TAG_RAISE;
rb_exc_raise(ec->errinfo);
}
Do not load file with same realpath twice when requiring This fixes issues with paths being loaded twice in certain cases when symlinks are used. It took me multiple attempts to get this working. My original attempt tried to convert paths to realpaths before adding them to $LOADED_FEATURES. Unfortunately, this doesn't work well with the loaded feature index, which is based off load paths and not realpaths. While I was able to get require working, I'm fairly sure the loaded feature index was not being used as expected, which would have significant performance implications. Additionally, I was never able to get that approach working with autoload when autoloading a non-realpath file. It also broke some specs. This takes a more conservative approach. Directly before loading the file, if the file with the same realpath has been required, the loading of the file is skipped. The realpaths are stored as fstrings in a hidden hash. When rebuilding the loaded feature index, the hash of realpaths is also rebuilt. I'm guessing this makes rebuilding process slower, but I don think that is a hot path. In general, modifying loaded features is only done when reloading, and that tends to be in non-production environments. Change test_require_with_loaded_features_pop test to use 30 threads and 300 iterations, instead of 4 threads and 1000 iterations. I saw only sporadic failures with 4/1000, but consistent failures 30/300 threads. These failures were due to the fact that the concurrent deletions from $LOADED_FEATURES in other threads can result in rb_ary_entry returning nil when rebuilding the loaded features index. To avoid concurrency issues when rebuilding the loaded features index, the building of the index itself is left alone, and afterwards, a separate loop is done on a copy of the loaded feature snapshot in order to rebuild the realpaths hash. Fixes [Bug #17885]
2021-06-30 23:50:19 +03:00
if (result == TAG_RETURN) {
2021-09-16 19:20:10 +03:00
rb_provide_feature(th2->vm, path);
VALUE real = realpath;
if (real) {
rb_hash_aset(realpaths, rb_fstring(real), Qtrue);
Do not load file with same realpath twice when requiring This fixes issues with paths being loaded twice in certain cases when symlinks are used. It took me multiple attempts to get this working. My original attempt tried to convert paths to realpaths before adding them to $LOADED_FEATURES. Unfortunately, this doesn't work well with the loaded feature index, which is based off load paths and not realpaths. While I was able to get require working, I'm fairly sure the loaded feature index was not being used as expected, which would have significant performance implications. Additionally, I was never able to get that approach working with autoload when autoloading a non-realpath file. It also broke some specs. This takes a more conservative approach. Directly before loading the file, if the file with the same realpath has been required, the loading of the file is skipped. The realpaths are stored as fstrings in a hidden hash. When rebuilding the loaded feature index, the hash of realpaths is also rebuilt. I'm guessing this makes rebuilding process slower, but I don think that is a hot path. In general, modifying loaded features is only done when reloading, and that tends to be in non-production environments. Change test_require_with_loaded_features_pop test to use 30 threads and 300 iterations, instead of 4 threads and 1000 iterations. I saw only sporadic failures with 4/1000, but consistent failures 30/300 threads. These failures were due to the fact that the concurrent deletions from $LOADED_FEATURES in other threads can result in rb_ary_entry returning nil when rebuilding the loaded features index. To avoid concurrency issues when rebuilding the loaded features index, the building of the index itself is left alone, and afterwards, a separate loop is done on a copy of the loaded feature snapshot in order to rebuild the realpaths hash. Fixes [Bug #17885]
2021-06-30 23:50:19 +03:00
}
}
2021-06-14 07:52:26 +03:00
ec->errinfo = saved.errinfo;
RUBY_DTRACE_HOOK(REQUIRE_RETURN, RSTRING_PTR(fname));
* probes.d: add DTrace probe declarations. [ruby-core:27448] * array.c (empty_ary_alloc, ary_new): added array create DTrace probe. * compile.c (rb_insns_name): allowing DTrace probes to access instruction sequence name. * Makefile.in: translate probes.d file to appropriate header file. * common.mk: declare dependencies on the DTrace header. * configure.in: add a test for existence of DTrace. * eval.c (setup_exception): add a probe for when an exception is raised. * gc.c: Add DTrace probes for mark begin and end, and sweep begin and end. * hash.c (empty_hash_alloc): Add a probe for hash allocation. * insns.def: Add probes for function entry and return. * internal.h: function declaration for compile.c change. * load.c (rb_f_load): add probes for `load` entry and exit, require entry and exit, and wrapping search_required for load path search. * object.c (rb_obj_alloc): added a probe for general object creation. * parse.y (yycompile0): added a probe around parse and compile phase. * string.c (empty_str_alloc, str_new): DTrace probes for string allocation. * test/dtrace/*: tests for DTrace probes. * vm.c (vm_invoke_proc): add probes for function return on exception raise, hash create, and instruction sequence execution. * vm_core.h: add probe declarations for function entry and exit. * vm_dump.c: add probes header file. * vm_eval.c (vm_call0_cfunc, vm_call0_cfunc_with_frame): add probe on function entry and return. * vm_exec.c: expose instruction number to instruction name function. * vm_insnshelper.c: add function entry and exit probes for cfunc methods. * vm_insnhelper.h: vm usage information is always collected, so uncomment the functions. 12 19:14:50 2012 Akinori MUSHA <knu@iDaemons.org> * configure.in (isinf, isnan): isinf() and isnan() are macros on DragonFly which cannot be found by AC_REPLACE_FUNCS(). This workaround enforces the fact that they exist on DragonFly. 12 15:59:38 2012 Shugo Maeda <shugo@ruby-lang.org> * vm_core.h (rb_call_info_t::refinements), compile.c (new_callinfo), vm_insnhelper.c (vm_search_method): revert r37616 because it's too slow. [ruby-dev:46477] * test/ruby/test_refinement.rb (test_inline_method_cache): skip the test until the bug is fixed efficiently. git-svn-id: svn+ssh://ci.ruby-lang.org/ruby/trunk@37631 b2dd03c8-39d4-4d8f-98ff-823fe69b080e
2012-11-13 01:52:12 +04:00
return result;
}
int
rb_require_internal_silent(VALUE fname)
{
rb_execution_context_t *ec = GET_EC();
return require_internal(ec, fname, 1, false);
}
2019-08-08 16:17:22 +03:00
int
rb_require_internal(VALUE fname)
2019-08-08 16:17:22 +03:00
{
rb_execution_context_t *ec = GET_EC();
return require_internal(ec, fname, 1, RTEST(ruby_verbose));
2019-08-08 16:17:22 +03:00
}
int
ruby_require_internal(const char *fname, unsigned int len)
{
struct RString fake;
VALUE str = rb_setup_fake_str(&fake, fname, len, 0);
2019-08-08 16:17:22 +03:00
rb_execution_context_t *ec = GET_EC();
int result = require_internal(ec, str, 0, RTEST(ruby_verbose));
rb_set_errinfo(Qnil);
return result == TAG_RETURN ? 1 : result ? -1 : 0;
}
VALUE
rb_require_string(VALUE fname)
{
2019-08-08 16:17:22 +03:00
rb_execution_context_t *ec = GET_EC();
int result = require_internal(ec, fname, 1, RTEST(ruby_verbose));
if (result > TAG_RETURN) {
2019-08-08 16:17:22 +03:00
EC_JUMP_TAG(ec, result);
}
if (result < 0) {
load_failed(fname);
}
2021-08-02 06:06:44 +03:00
return RBOOL(result);
}
VALUE
rb_require(const char *fname)
{
return rb_require_string(rb_str_new_cstr(fname));
}
static int
register_init_ext(st_data_t *key, st_data_t *value, st_data_t init, int existing)
{
const char *name = (char *)*key;
if (existing) {
/* already registered */
rb_warn("%s is already registered", name);
}
else {
*value = (st_data_t)MEMO_NEW(0, 0, init);
*key = (st_data_t)ruby_strdup(name);
}
return ST_CONTINUE;
}
RUBY_FUNC_EXPORTED void
ruby_init_ext(const char *name, void (*init)(void))
{
2021-09-16 19:20:10 +03:00
rb_vm_t *vm = GET_VM();
st_table *loading_tbl = get_loading_table(vm);
2021-09-16 19:20:10 +03:00
if (feature_provided(vm, name, 0))
return;
st_update(loading_tbl, (st_data_t)name, register_init_ext, (st_data_t)init);
}
/*
* call-seq:
* mod.autoload(const, filename) -> nil
*
* Registers _filename_ to be loaded (using Kernel::require)
* the first time that _const_ (which may be a String or
* a symbol) is accessed in the namespace of _mod_.
*
* module A
* end
* A.autoload(:B, "b")
* A::B.doit # autoloads "b"
*
* If _const_ in _mod_ is defined as autoload, the file name to be
* loaded is replaced with _filename_. If _const_ is defined but not
* as autoload, does nothing.
*/
static VALUE
rb_mod_autoload(VALUE mod, VALUE sym, VALUE file)
{
ID id = rb_to_id(sym);
FilePathValue(file);
rb_autoload_str(mod, id, file);
return Qnil;
}
/*
* call-seq:
* mod.autoload?(name, inherit=true) -> String or nil
*
* Returns _filename_ to be loaded if _name_ is registered as
* +autoload+ in the namespace of _mod_ or one of its ancestors.
*
* module A
* end
* A.autoload(:B, "b")
* A.autoload?(:B) #=> "b"
*
* If +inherit+ is false, the lookup only checks the autoloads in the receiver:
*
* class A
* autoload :CONST, "const.rb"
* end
*
* class B < A
* end
*
* B.autoload?(:CONST) #=> "const.rb", found in A (ancestor)
* B.autoload?(:CONST, false) #=> nil, not found in B itself
*
*/
static VALUE
rb_mod_autoload_p(int argc, VALUE *argv, VALUE mod)
{
int recur = (rb_check_arity(argc, 1, 2) == 1) ? TRUE : RTEST(argv[1]);
VALUE sym = argv[0];
ID id = rb_check_id(&sym);
if (!id) {
return Qnil;
}
return rb_autoload_at_p(mod, id, recur);
}
/*
* call-seq:
* autoload(const, filename) -> nil
*
* Registers _filename_ to be loaded (using Kernel::require)
* the first time that _const_ (which may be a String or
* a symbol) is accessed.
*
* autoload(:MyModule, "/usr/local/lib/modules/my_module.rb")
*
* If _const_ is defined as autoload, the file name to be loaded is
* replaced with _filename_. If _const_ is defined but not as
* autoload, does nothing.
*/
static VALUE
rb_f_autoload(VALUE obj, VALUE sym, VALUE file)
{
VALUE klass = rb_class_real(rb_vm_cbase());
2020-10-08 19:21:10 +03:00
if (!klass) {
rb_raise(rb_eTypeError, "Can not set autoload on singleton class");
}
return rb_mod_autoload(klass, sym, file);
}
/*
* call-seq:
* autoload?(name, inherit=true) -> String or nil
*
* Returns _filename_ to be loaded if _name_ is registered as
* +autoload+.
*
* autoload(:B, "b")
* autoload?(:B) #=> "b"
*/
static VALUE
rb_f_autoload_p(int argc, VALUE *argv, VALUE obj)
{
/* use rb_vm_cbase() as same as rb_f_autoload. */
VALUE klass = rb_vm_cbase();
if (NIL_P(klass)) {
return Qnil;
}
return rb_mod_autoload_p(argc, argv, klass);
}
void
Init_load(void)
{
rb_vm_t *vm = GET_VM();
static const char var_load_path[] = "$:";
ID id_load_path = rb_intern2(var_load_path, sizeof(var_load_path)-1);
rb_define_hooked_variable(var_load_path, (VALUE*)vm, load_path_getter, rb_gvar_readonly_setter);
rb_alias_variable(rb_intern_const("$-I"), id_load_path);
rb_alias_variable(rb_intern_const("$LOAD_PATH"), id_load_path);
vm->load_path = rb_ary_new();
vm->expanded_load_path = rb_ary_hidden_new(0);
vm->load_path_snapshot = rb_ary_hidden_new(0);
Fix compatibility of cached expanded load path * file.c (rb_get_path_check_to_string): extract from rb_get_path_check(). We change the spec not to call to_path of String object. * file.c (rb_get_path_check_convert): extract from rb_get_path_check(). * file.c (rb_get_path_check): follow the above change. * file.c (rb_file_expand_path_fast): remove check_expand_path_args(). Instead we call it in load.c. * file.c (rb_find_file_ext_safe): use rb_get_expanded_load_path() to reduce expand cost. * file.c (rb_find_file_safe): ditto. * internal.h (rb_get_expanded_load_path): add a declaration. * internal.h (rb_get_path_check_to_string, rb_get_path_check_convert): add declarations. * load.c (rb_construct_expanded_load_path): fix for compatibility. Same checks in rb_get_path_check() are added. We don't replace $LOAD_PATH and ensure that String object of $LOAD_PATH are frozen. We don't freeze non String object and expand it every times. We add arguments for expanding load path partially and checking if load path have relative paths or non String objects. * load.c (load_path_getcwd): get current working directory for checking if it's changed when getting load path. * load.c (rb_get_expanded_load_path): fix for rebuilding cache properly. We check if current working directory is changed and rebuild expanded load path cache. We expand paths which start with ~ (User HOME) and non String objects every times for compatibility. We make this accessible from other source files. * load.c (rb_feature_provided): call rb_get_path() since we changed rb_file_expand_path_fast() not to call it. * load.c (Init_load): initialize vm->load_path_check_cache. * vm.c (rb_vm_mark): mark vm->load_path_check_cache for GC. * vm_core.h (rb_vm_struct): add vm->load_path_check_cache to store data to check load path cache validity. * test/ruby/test_require.rb (TestRequire): add tests for require compatibility related to cached expanded load path. [ruby-core:47970] [Bug #7158] git-svn-id: svn+ssh://ci.ruby-lang.org/ruby/trunk@37482 b2dd03c8-39d4-4d8f-98ff-823fe69b080e
2012-11-05 19:27:08 +04:00
vm->load_path_check_cache = 0;
rb_define_singleton_method(vm->load_path, "resolve_feature_path", rb_resolve_feature_path, 1);
rb_define_virtual_variable("$\"", get_LOADED_FEATURES, 0);
rb_define_virtual_variable("$LOADED_FEATURES", get_LOADED_FEATURES, 0);
vm->loaded_features = rb_ary_new();
vm->loaded_features_snapshot = rb_ary_hidden_new(0);
vm->loaded_features_index = st_init_numtable();
Do not load file with same realpath twice when requiring This fixes issues with paths being loaded twice in certain cases when symlinks are used. It took me multiple attempts to get this working. My original attempt tried to convert paths to realpaths before adding them to $LOADED_FEATURES. Unfortunately, this doesn't work well with the loaded feature index, which is based off load paths and not realpaths. While I was able to get require working, I'm fairly sure the loaded feature index was not being used as expected, which would have significant performance implications. Additionally, I was never able to get that approach working with autoload when autoloading a non-realpath file. It also broke some specs. This takes a more conservative approach. Directly before loading the file, if the file with the same realpath has been required, the loading of the file is skipped. The realpaths are stored as fstrings in a hidden hash. When rebuilding the loaded feature index, the hash of realpaths is also rebuilt. I'm guessing this makes rebuilding process slower, but I don think that is a hot path. In general, modifying loaded features is only done when reloading, and that tends to be in non-production environments. Change test_require_with_loaded_features_pop test to use 30 threads and 300 iterations, instead of 4 threads and 1000 iterations. I saw only sporadic failures with 4/1000, but consistent failures 30/300 threads. These failures were due to the fact that the concurrent deletions from $LOADED_FEATURES in other threads can result in rb_ary_entry returning nil when rebuilding the loaded features index. To avoid concurrency issues when rebuilding the loaded features index, the building of the index itself is left alone, and afterwards, a separate loop is done on a copy of the loaded feature snapshot in order to rebuild the realpaths hash. Fixes [Bug #17885]
2021-06-30 23:50:19 +03:00
vm->loaded_features_realpaths = rb_hash_new();
rb_obj_hide(vm->loaded_features_realpaths);
rb_define_global_function("load", rb_f_load, -1);
rb_define_global_function("require", rb_f_require, 1);
rb_define_global_function("require_relative", rb_f_require_relative, 1);
rb_define_method(rb_cModule, "autoload", rb_mod_autoload, 2);
rb_define_method(rb_cModule, "autoload?", rb_mod_autoload_p, -1);
rb_define_global_function("autoload", rb_f_autoload, 2);
rb_define_global_function("autoload?", rb_f_autoload_p, -1);
ruby_dln_librefs = rb_ary_hidden_new(0);
rb_gc_register_mark_object(ruby_dln_librefs);
}