USRMARSHAL. [ruby-dev:21582]
* marshal.c (w_object): dump generic instance variables from
a string from '_dump'.
* variable.c (rb_generic_ivar_table): return 0 if obj's FL_EXIVAR
is not set.
* time.c (time_dump): copy instance variables to dumped string, to
be included in the marshaled data.
* bignum.c (rb_big2ulong): add range check to ensure round trip.
git-svn-id: svn+ssh://ci.ruby-lang.org/ruby/trunk@4734 b2dd03c8-39d4-4d8f-98ff-823fe69b080e
negative values are not "too big". [ruby-dev21567]
* marshal.c (w_object): should pass the value of "weak" to
w_object(). [ruby-dev:21555] and [ruby-dev:21561]
git-svn-id: svn+ssh://ci.ruby-lang.org/ruby/trunk@4729 b2dd03c8-39d4-4d8f-98ff-823fe69b080e
loading from marshaled data.
* time.c (time_mdump): new marshal dumper. _dump is still
available for compatibility.
* time.c (time_mload): new marshal loader.
* marshal.c (w_object): preserve instance variables for objects
with marshal_dump.
* marshal.c (r_object0): restore instance variables before calling
marshal_load.
* error.c (rb_warn_m): always return nil.
git-svn-id: svn+ssh://ci.ruby-lang.org/ruby/trunk@4651 b2dd03c8-39d4-4d8f-98ff-823fe69b080e
before detach pid. [ruby-talk:71519]
* eval.c (PUSH_FRAME): save outer ruby_block. [ruby-list:37677],
[ruby-dev:20202]
* eval.c (BEGIN_CALLARGS): restore outer block by using
ruby_block->outer.
* eval.c (block_pass): do not alter block->prev, but block->outer.
* array.c (get_inspect_tbl): warning on wrong condition.
* eval.c (localjump_xvalue): renamed exitstatus to exit_value
since it's not exit "status" after all.
* eval.c (localjump_error): add reason to LocalJumpError.
* compar.c (rb_cmpint): raise error via rb_cmperr(), if cmp value
is nil. now take new 2 arguments.
* time.c (time_cmp): 2003-05-16 fix was incomplete.
(ruby-bugs-ja:PR#458)
git-svn-id: svn+ssh://ci.ruby-lang.org/ruby/trunk@3823 b2dd03c8-39d4-4d8f-98ff-823fe69b080e
since it works as copy constructor.
* eval.c (rb_add_method): initialize_copy should always be
private, like initialize.
git-svn-id: svn+ssh://ci.ruby-lang.org/ruby/trunk@3820 b2dd03c8-39d4-4d8f-98ff-823fe69b080e
* time.c (time_cmp): does not compare with numbers for
interchangeability. (ruby-bugs-ja PR#458)
git-svn-id: svn+ssh://ci.ruby-lang.org/ruby/trunk@3808 b2dd03c8-39d4-4d8f-98ff-823fe69b080e
* parse.y (negate_lit): new function to negate literal numeric
values in compile time.
* regex.c (re_match_exec): charset info may be stored in MBC
region when $KCODE != NONE.
* error.c (set_syserr): should preserve duplicated error names.
git-svn-id: svn+ssh://ci.ruby-lang.org/ruby/trunk@3397 b2dd03c8-39d4-4d8f-98ff-823fe69b080e
* parse.y (cmd_brace_block): new tLBRACE_ARG block rule
* parse.y (command): can take optional cmd_brace_block; use %prec
to resolve shift/reduce conflict. (ruby-bugs-ja PR#372)
* eval.c (ruby_finalize): trace_func should be cleared here (after
executing exit procs and finalizers).
* eval.c (rb_define_alloc_func): new allocation framework, based
on Nobu's work [ruby-dev:19116]. "allocate" method is no longer
used for object allocation.
git-svn-id: svn+ssh://ci.ruby-lang.org/ruby/trunk@3188 b2dd03c8-39d4-4d8f-98ff-823fe69b080e
still left; removed.
* sprintf.c (rb_f_sprintf): should not prepend '0' if width > prec
for example "%5.3d".
* process.c (Init_process): add Process.exit and Process.abort
* pack.c (utf8_to_uv): raise ArgumentError for malformed/redundant
UTF-8 sequences.
* process.c (last_status_set): add pid attribute to Process::Status.
* pack.c (uv_to_utf8): limit maximum length of the encoded string
to 6 bytes, even when the platform supports 8 bytes long integers.
* pack.c (utf8_to_uv): do not decode sequences longer than 6 bytes.
* object.c (copy_object): use "copy_object" method, not "become".
git-svn-id: svn+ssh://ci.ruby-lang.org/ruby/trunk@3123 b2dd03c8-39d4-4d8f-98ff-823fe69b080e
empty path, just ignore.
* dir.c: remove <sys/parm.h> inclusion.
* compar.c (cmp_eq,cmp_gt,cmp_ge,cmp_lt,cmp_le): check using
rb_cmpint().
* error.c (init_syserr): remove sys_nerr dependency.
* numeric.c (num_cmp): added to satisfy Comparable assumption.
* eval.c (rb_add_method): "initialize" should be public if it is a
singleton method.
* regex.c (re_match): avoid dereferencing if size == 0.
(ruby-bugs-ja:PR#360)
* time.c (time_cmp): should return nil if an operand is not a
number nor time. (ruby-bugs-ja:PR#359)
* file.c (rb_stat_cmp): should return nil if an operand is not
File::Stat.
git-svn-id: svn+ssh://ci.ruby-lang.org/ruby/trunk@3076 b2dd03c8-39d4-4d8f-98ff-823fe69b080e
* win32/win32.c (rb_w32_stat): remove S_IWGRP and S_IWOTH bits from
st_mode.
* win32/win32.h (S_I*): define if not defined.
git-svn-id: svn+ssh://ci.ruby-lang.org/ruby/trunk@2822 b2dd03c8-39d4-4d8f-98ff-823fe69b080e
variable table if it existes.
* class.c (rb_make_metaclass): metaclass of a metaclass is a
metaclass itself.
git-svn-id: svn+ssh://ci.ruby-lang.org/ruby/trunk@2784 b2dd03c8-39d4-4d8f-98ff-823fe69b080e
* file.c (rb_find_file_ext): ditto.
* range.c (range_eq): class check should be based on range.class,
instead of Range to work with Range.dup.
* range.c (range_eql): ditto.
* class.c (rb_mod_dup): need to preserve metaclass and flags.
* object.c (rb_cstr_to_dbl): had a buffer overrun.
* marshal.c (w_class): integrate singleton check into a funciton
to follow DRY principle.
* marshal.c (w_uclass): should check singleton method.
* object.c (rb_obj_dup): dmark and dfree functions must be match
for T_DATA type.
* object.c (rb_obj_dup): class of the duped object must be match
to the class of the original.
* re.c (rb_reg_quote): do not escape \t, \f, \r, \n, for they are
not regular expression metacharacters.
* time.c (time_s_alloc): use time_free instead of free (null check,
also serves for type mark).
* time.c (time_s_at): check dfree function too.
git-svn-id: svn+ssh://ci.ruby-lang.org/ruby/trunk@2748 b2dd03c8-39d4-4d8f-98ff-823fe69b080e
Numeric. Otherwise using "succ".
* range.c (range_each): iteration done using "succ". If the
elements does not respond to "succ", raise TypeError. As a
result, all Enumerable methods, e.g. collect, require elements
to respond to "succ'.
* range.c (range_member): comparison done using "each", if
elements are non-Numeric or no-"succ" objects. Otherwise
compare using "<=>".
* range.c (Init_Range): remove "size" and "length".
git-svn-id: svn+ssh://ci.ruby-lang.org/ruby/trunk@2506 b2dd03c8-39d4-4d8f-98ff-823fe69b080e
* ruby.c (proc_options): access prefixed "ruby_yydebug".
* applied modifies to pacify some of gcc -Wall warnings.
* parse.y (arg): no more ugly hack for "**", so that "-2**2" to be
parsed as "(-2)**2", whereas "- 2**2" or "-(2)**2" to be parsed
as "-(2**2)".
* parse.y (yylex): '-2' to be literal fixnum. [new]
* time.c (time_succ): new method for Range support.
* time.c (time_arg): nil test against v[6] (usec).
git-svn-id: svn+ssh://ci.ruby-lang.org/ruby/trunk@2500 b2dd03c8-39d4-4d8f-98ff-823fe69b080e
* io.c (rb_io_mode_modenum): ditto.
* gc.c (rb_memerror): rename from mem_error, and exported.
* gc.c (Init_GC): pre-allocate NoMemoryError instance.
* object.c (convert_type): error message changed from "failed to
convert" to "cannot convert", since it does not try to convert
if an object does not respond to the converting method.
* eval.c (block_pass): convert Method to Proc using
rb_check_convert_type().
* object.c (rb_check_convert_type): always convert T_DATA
* eval.c (rb_thread_cleanup): should not terminate main_thread by
Fatal error.
* regex.c (is_in_list): need to not exclude NUL and NEWLINE.
* re.c (rb_reg_expr_str): wrong backslash escapement.
* re.c (rb_reg_expr_str): do not escape embedded space
characters.
* marshal.c (w_object): T_DATA process patch from Joel VanderWerf
<vjoel@PATH.Berkeley.EDU>. This is temporary hack; it remains
undocumented, and it will be removed when marshaling is
re-designed.
* marshal.c (r_object): ditto.
* numeric.c (num_step): Integer#step is moved to Numeric#step;
Fixnum#step is merged into this method.
* numeric.c (int_dotimes): Fixnum#times is merged.
* numeric.c (int_upto): Fixnum#upto is merged.
* numeric.c (int_downto): Fixnum#downto is merged.
git-svn-id: svn+ssh://ci.ruby-lang.org/ruby/trunk@2401 b2dd03c8-39d4-4d8f-98ff-823fe69b080e
<Masamichi.Akiyoshi@jp.compaq.com> is merged.
* eval.c (exec_under): changing ruby_class is OK, but should not
alter cbase.
* eval.c (yield_under_i): ditto.
git-svn-id: svn+ssh://ci.ruby-lang.org/ruby/trunk@2258 b2dd03c8-39d4-4d8f-98ff-823fe69b080e
singleton cbase.
* eval.c (is_defined): adopt new cvar behavior.
* eval.c (rb_eval): ditto.
* eval.c (assign): ditto.
* class.c (rb_mod_clone): should not call rb_obj_clone(), since
Module does not provide "allocate".
* class.c (rb_singleton_class): should crate new singleton class
if obj is a class or module and attached object is different,
which means metaclass of singleton class is sought.
* time.c (time_s_alloc): now follows allocation framework.
* eval.c (rb_eval): should initialize outer class variables from
methods in singleton class definitions.
* eval.c (assign): ditto.
git-svn-id: svn+ssh://ci.ruby-lang.org/ruby/trunk@2169 b2dd03c8-39d4-4d8f-98ff-823fe69b080e
* eval.c (rb_thread_schedule): should not select a thread which is
not yet initialized.
* time.c (time_plus): wrong boundary check.
* time.c (time_minus): ditto.
git-svn-id: svn+ssh://ci.ruby-lang.org/ruby/trunk@1918 b2dd03c8-39d4-4d8f-98ff-823fe69b080e
(time_utc_offset): new function.
(Init_Time): bind above methods to `time_utc_offset'.
* time.c: 64bit time_t support.
(time_s_at): use NUM2LONG instead of NUM2INT for tv_sec.
(time_arg): initialize tm_isdst correctly.
use long to initialize tm_year.
(search_time_t): renamed from `make_time_t'.
(make_time_t): call `timegm' and `mktime' instead of `search_time_t'
if availabe.
(time_to_i): use LONG2NUM instead of INT2NUM.
(time_localtime): check localtime failure.
(time_gmtime): check gmtime failure.
(time_year): use LONG2NUM instead of INT2FIX.
(time_to_a): use long for tm_year.
(time_dump): check tm_year which is not representable with 17bit.
(time_load): initialize tm_isdst.
* configure.in: check existence of `mktime' and `timegm'.
check existence of tm_gmtoff field of struct tm.
fix negative time_t for 64bit time_t.
* missing/strftime.c: fix overflow by tm_year + 1900.
git-svn-id: svn+ssh://ci.ruby-lang.org/ruby/trunk@1912 b2dd03c8-39d4-4d8f-98ff-823fe69b080e
div, mod.
* time.c (time_cmp): should handle Bignums.
* array.c (rb_ary_pop): should ELTS_SHARED flag check before
REALLOC.
git-svn-id: svn+ssh://ci.ruby-lang.org/ruby/trunk@1909 b2dd03c8-39d4-4d8f-98ff-823fe69b080e
modifying buffer is shared.
* array.c (ary_make_shared): make an internal buffer of an array
to be shared.
* array.c (rb_ary_shift): avoid sliding an internal buffer by
using shared buffer.
* array.c (rb_ary_subseq): avoid copying the buffer.
* parse.y (gettable): should freeze __LINE__ string.
* io.c (rb_io_puts): old behavoir restored. rationale: a) if you
want to call to_s for arrays, you can just call print a, "\n".
b) to_s wastes memory if array (and sum of its contents) is
huge. c) now any object that has to_ary is treated as an array,
using rb_check_convert_type().
* hash.c (rb_hash_initialize): now accepts a block to calculate
the default value. [new]
* hash.c (rb_hash_aref): call "default" method to get the value
corrensponding to the non existing key.
* hash.c (rb_hash_default): get the default value based on the
block given to 'new'. Now it takes an optinal "key" argument.
"default" became the method to get the value for non existing
key. Users may override "default" method to change the hash
behavior.
* hash.c (rb_hash_set_default): clear the flag if a block is given
to 'new'
* object.c (Init_Object): undef Data.allocate, left Data.new.
* ext/curses/curses.c (window_scrollok): use RTEST().
* ext/curses/curses.c (window_idlok): ditto.
* ext/curses/curses.c (window_keypad): ditto.
* ext/curses/curses.c (window_idlok): idlok() may return void on
some platforms; so don't use return value.
* ext/curses/curses.c (window_scrollok): ditto for consistency.
* ext/curses/curses.c: replace FIX2INT() by typechecking NUM2INT().
* parse.y (str_extend): should not process immature #$x and
#@x interpolation, e.g #@#@ etc.
* enum.c (enum_sort_by): sort_by does not have to be stable always.
* enum.c (enum_sort_by): call qsort directly to gain performance.
* util.c (ruby_qsort): ruby_qsort(qs6) is now native thread safe.
* error.c (rb_sys_fail): it must be a bug if it's called when
errno == 0.
* regex.c (WC2MBC1ST): should not pass through > 0x80 number in UTF-8.
git-svn-id: svn+ssh://ci.ruby-lang.org/ruby/trunk@1896 b2dd03c8-39d4-4d8f-98ff-823fe69b080e
nobu
siena
dave
matz
akr
michal
usa
eban