The order of iseq may differ from the order of tokens, typically
`while`/`until` conditions are put after the body.
These orders can match by using line numbers as builtin-indexes, but
at the same time, it introduces the restriction that multiple `cexpr!`
and `cstmt!` cannot appear in the same line.
Another possible idea is to use `RubyVM::AbstractSyntaxTree` and
`node_id` instead of ripper, with making BASERUBY 3.1 or later.
Before this commit the Prism compiler would try to intern constants
every time it re-entered. This pool of constants is "constant" (there is
only one pool per parser instance), so we should do it only once: upon
the top level entry to the compiler.
This change does just that: it populates the interned constants once.
Fixes: https://github.com/ruby/prism/issues/2152
Previously on builds with optimizations disabled, this could result in
an out of bounds read. When we had all of:
* built with -O0
* Leaf builtin
* Primitive.mandatory_only
* "no args builtin", called by vm_call_single_noarg_inline_builti
* The stack is escaped to the heap via binding or a proc
This is because mk_builtin_loader generated reads for all locals
regardless of whether they were used and in the case we generated a
mandatory_only iseq that would include more variables than were actually
available.
On optimized builds, the invalid accesses would be optimized away, and
this also was often unnoticed as the invalid access would just hit
another part of the stack unless it had been escaped to the heap.
The fix here is imperfect, as this could have false positives, but since
Primitive.cexpr! is only available within the cruby codebase itself
that's probably fine as a proper fix would be much more challenging (the
only false positives we found were in rjit.rb).
Fixes [Bug #20178]
Co-authored-by: Adam Hess <HParker@github.com>
`:sym` was managed by `NODE_LIT` with `Symbol` object.
This commit introduces `NODE_SYM` so that
1. Symbol literal is detectable from AST Node
2. Reduce dependency on ruby object
RDoc options that do not change and can be written in `.rdoc_options`
file are moved, so that they match when called without `make`.
Get rid of parsing the files in `page_dir` twice (as relative paths
and absolute paths).
I'm almost certain nobody is actually running this script on Windows,
but the tests for it do run during `nmake check`, and they fail at least
on my git configuration.
The $ anchor doesn't match \r\n with git's -E regex matching, so we need
to add \r? to gobble the carriage-return up too if needed.
The test is currently skipped and can't possibly work on windows at the
moment. It fails because $LOAD_PATH is not set up properly in the forked
CGI process, so `require 'uri'` fails.
This works properly in the test_cgi.rb tests, because it sets up a
:RequestCallback to fix things up. Let's move the setup there into
util.rb, so it can be shared with test_filehandler.rb as well.
When something went wrong and git launches editor, and a editor chosen
was terminal-based, a test run unnoticeably hangs.
As we intend editors not to be ran with --no-edit, GIT_EDITOR should be
defaulted to `false` so let Git command fails when it attempts to start a
editor. This allows catching such unintentional behaviour in test suite.
(Note: Prior to Git v2.32.0, git cherry-pick --no-edit doesn't work for
conflict resolution so currently the test hangs when older Git is in use.
39edfd5cbc
i.e. Ubuntu 20.04, Debian Bullseye doesn't satisfy this prerequisite.)
Previously, the embedded semicolon in BASERUBY if BASERUBY is
not available breaks tarball builds without BASERUBY when using
OpenBSD make, due to the inability to escape MFLAGS correctly.
This moves the same BASERUBY code into a separate file, avoiding
the MFLAGS quoting issue.
BASERUBY must be passed to build-ext because it is required
by ripper since the introduction of lrama.
Fixes [Bug #19683]
Co-authored-by: Nobuyoshi Nakada <nobu@ruby-lang.org>
This commit adds `GC.auto_compact = :empty` which will run
auto-compaction sorting pages by empty slots so the most amount of
objects will be moved. This will make it easier to write tests for
auto-compaction.
Some specs now print the following warning:
```
/path/to/bundler/tmp/1/gems/base/ruby/3.2.0/gems/sinatra-2.0.8.1/lib/sinatra/base.rb:902: warning: constant Tilt::Cache is deprecated
```
Updating sinatra to latest & greatest fixes it.
Update other deps too since at it.
This makes bundler consistent with all other gems, and makes the default
installation of Bundler in the release package look like any other
bundler installation.
Before (on preview3, for example), Bundler executable is installed at:
lib/ruby/gems/3.3.0+0/gems/bundler-2.5.0.dev/libexec/bundle
Now it's installed in the standard location:
lib/ruby/gems/3.3.0+0/gems/bundler-2.5.0.dev/exe/bundle
```
.../tool/lib/colorize.rb:56:in `resolve_color': undefined method `gsub' for an instance of Symbol (NoMethodError)
color.gsub(/\b[a-z][\w ]+/) do |n|
^^^^^
from .../tool/lib/colorize.rb:47:in `decorate'
from .../tool/lib/test/unit.rb:1012:in `block in failed'
```
* Provide a new API compile_file_prism which mirrors compile_file
but uses prism to parse/compile.
* Provide the ability to run test-all with RUBY_ISEQ_DUMP_DEBUG set
to "prism". If it is, we'll use the new compile_file_prism API to
load iseqs during the test run.
test/runner.rb and tool/lib/test/unit/parallel.rb must use the same
settings. However, some settings were copied and pasted, while some were
added only to test/runner.rb.
This changeset creates tool/test/init.rb for all settings of test-unit,
which is loaded not only by test/runner.rb but also
tool/lib/test/unit/parallel.rb.
Background: the GEM_HOME environment variable was removed in
test/runner.rb, which prohibit `require "rake"` (note that rake is a
bundled gem). However the parallel mode didn't refrect this setting,
i.e., `require "rake"` was allowed.
This leads to an inconsistency, which actually affected a test test
defines s test class *only when* `require "rake"` is successful.
(test/rubygems/test_gem_package_task.rb)
https://github.com/ruby/ruby/actions/runs/6807729617/job/18511055636#step:8:1714
```
/home/runner/work/ruby/ruby/src/tool/lib/test/unit.rb:729:in `const_get': uninitialized constant TestGemPackageTask (NameError)
suites.map! {|r| ::Object.const_get(r[:testcase])}
^^^^^^^^^^
```
* Remove CI functions for make test-all.
I want to simplify the `.travis.yml` for us to maintain it easily. When we
need to skip a test, we can fix the test file directly.
I removed the `tool/ci_functions.sh` too. I confirmed the file is not used in
any other files.
* .travis.yml: Remove a notification for IRC channel.
GitHub Actions don't use this notification. The setting is outdated.
* Fix to skip the commit including the document keyword.
There were 2 `if` syntax lines in the file.
* Remove ccache.
to @__passed__, @@__current__.
@passed is redefined in a few test suites, and this could lead to bugs.
Also rename @options (Runner#options) to @__runner_options__, which is
only used in make test-tool anyway.
This reverts the commits for the master branch of RBS:
- commit f717faac632dd8664d0967f8e639b84d5d032854: "Update rbs
revision to test"
The target revision to test is in master branch, not for 3.2.x.
- commit 9e93af5329f35092c3de3ea37d4e9e181b800bb2: "Skip RBS
`RbConfig::TOPDIR` test that is `nil` before installation"
RbConfig_test.rb is not updated in 3.2.x branch.
I think it meant it's now a bundled "gem", but "is now bundled" implies
it's a new gem that was neither default nor bundled gems. So I changed
wording.
I also want to change the subjects and reorder them so that it will read:
"The following default gem" (new) -> "The following default gem" (updated) ->
"The following bundled gem" (new) -> "The following bundled gem" (updated).
On some platforms, such as FreeBSD and Oracle Linux, symbols defined
in the crt0 setup routine are exported from shared libraries. So
ignore the symbols that would be exported even in an empty shared
library.
* Port call threshold logic from Rust to C for performance
* Prefix global/field names with yjit_
* Fix linker error
* Fix preprocessor condition for rb_yjit_threshold_hit
* Fix third linker issue
* Exclude yjit_calls_at_interv from RJIT bindgen
---------
Co-authored-by: Takashi Kokubun <takashikkbn@gmail.com>
The case of 7fc73ab5f6, which modified
`.gitignore` and `.github/workflows/main.yml`. Both files need to be
rejected and restored, but since the latter file was not there before,
`git checkout` failed and the former file could not be restored along
with it. To fix this failure, restore the ignored files one by one.
Yet another partial revert of https://github.com/ruby/ruby/pull/8329,
similar to 00f263e6c4.
Repro: On ruby/ruby@1be64e34d0, `tool/sync_default_gems.rb yarp
162c2088eec6ec8f0558559e082cd661c18ee02a` should exist successfully, but
it doesn't without this gem-specific handling.
- Filter out files to be ignored first, then resolve conflicts.
- Add "added by gem" files, instead of hard-code paths to add.
- Remove gem specific patterns covered by more generic rules.
See bug https://bugs.freebsd.org/bugzilla/show_bug.cgi?id=271490
On FreeBSDk, it's possible for fork(2) in a multithreaded process to
hang because of a bug in the lock handling of the dynamic linker. This
is now fixed on FreeBSD master, but it would be good if we could work
around it for Ruby CI which is running 13.1.
Setting LD_BIND_NOW seems to work around the problem (probably because
the dynamic linker doesn't then need to resolve anything through the PLT
when it's first called).
* Add yarp/yarp_compiler.c as stencil for compiling YARP
This commit adds yarp/yarp_compiler.c, and changes the sync script
to ensure that yarp/yarp_compiler.c will not get overwritten
* [Misc #119772] Create and expose RubyVM::InstructionSequence.compile_yarp
This commit creates the stencil for a compile_yarp function, which
we will continue to fill out. It allows us to check the output
of compiled YARP code against compiled code without using YARP.
I was testing this script on git.ruby-lang.org to use its git version,
but it did not work because the server's default user doesn't have SSH
keys.
https works for everyone, so it's a safer default. You shouldn't need to
push to that remote from ruby/ruby anyway.
Nobuyoshi Nakada
Hiroshi SHIBATA
Kevin Newton
eileencodes
David Rodríguez
Kazuhiro NISHIYAMA
Takashi Kokubun
yui-knk
Aaron Patterson
John Hawthorn
S-H-GAMELINKS
Luke Gruber
KJ Tsanaktsidis
Sorah Fukumori
Jeremy Evans
Peter Zhu
Samuel Giddins
git
NARUSE, Yui
Soutaro Matsumoto
Yusuke Endoh
Soutaro Matsumoto
Jun Aruga
Maxime Chevalier-Boisvert
KJ Tsanaktsidis
Jemma Issroff
Samuel Williams