AIX. [ruby-talk:61466]
* marshal.c (math.h): should be included after ruby.h on AIX.
[ruby-talk:61366]
git-svn-id: svn+ssh://ci.ruby-lang.org/ruby/trunk@3342 b2dd03c8-39d4-4d8f-98ff-823fe69b080e
classes are not allowed inside character classes.
[ruby-talk:60996]
git-svn-id: svn+ssh://ci.ruby-lang.org/ruby/trunk@3325 b2dd03c8-39d4-4d8f-98ff-823fe69b080e
fflush() are needed.
* io.c (flush_before_seek): flush write stream only.
* io.c (rb_io_check_readable): seek instead of flush if the last
operation was write.
* io.c (rb_io_check_writable): seek instead of flush if the last
operation was read.
* bcc32/Makefile.sub, win32/Makefile.sub: needs to seek between
R/W.
git-svn-id: svn+ssh://ci.ruby-lang.org/ruby/trunk@3318 b2dd03c8-39d4-4d8f-98ff-823fe69b080e
unnecessary
* intern.h: debug does not run if ID_ALLOCATOR is zero.
git-svn-id: svn+ssh://ci.ruby-lang.org/ruby/trunk@3310 b2dd03c8-39d4-4d8f-98ff-823fe69b080e
* hash.c (env_shift): ditto.
* hash.c (env_invert): ditto.
* hash.c (env_replace): ditto.
* hash.c (env_update): ditto.
* array.c (rb_ary_join): dispatch based on "to_str".
* array.c (rb_ary_times): ditto.
* array.c (rb_ary_equal): ditto.
* process.c (rb_f_exec): dispatch based on "to_ary".
* eval.c (umethod_bind): exact class match is not required. relax
the restriction to subclasses.
* eval.c (rb_eval): call "inherited" before executing class body.
* class.c (rb_define_class): call "inherited" after defining the
constant.
* class.c (rb_define_class_under): ditto.
* eval.c (massign): expand first element if RHS is an array and
its size is 1, and LHS has concrete assignment target (i.e. LHS
has target(s) other than *var).
* eval.c (massign): avoid unnecessary avalue/svalue conversion.
* eval.c (rb_yield_0): ditto
* array.c (rb_ary_update): do not allocate unused array if rpl is
nil (i.e. merely removing elements).
* io.c (io_read): should resize supplied string if it's shorter
than expected.
git-svn-id: svn+ssh://ci.ruby-lang.org/ruby/trunk@3307 b2dd03c8-39d4-4d8f-98ff-823fe69b080e
I was wrong about something. It was a return value,
not arguments that are not covered by the int-to-long promotion rule
on IA64, if the function is not explicitly declared.
The commit itself was generally a right thing to do, but just let me
correct my commit message.
git-svn-id: svn+ssh://ci.ruby-lang.org/ruby/trunk@3304 b2dd03c8-39d4-4d8f-98ff-823fe69b080e
st_compare_func_t, st_hash_func_t and st_each_func_t.
* st.h, st.c: Do explicit function declarations and do not rely on
implicit declarations. On such platforms as IA64, int argument
values are NOT automatically promoted to long (64bit) values, so
explicit declarations are mandatory for those functions that
take long values or pointers. This fixes miniruby's coredump on
FreeBSD/IA64.
* class.c, eval.c, gc.c, hash.c, marshal.c, parse.y, variable.c:
Add proper casts to avoid warnings.
git-svn-id: svn+ssh://ci.ruby-lang.org/ruby/trunk@3303 b2dd03c8-39d4-4d8f-98ff-823fe69b080e
* ext/digest/md5/md5ossl.c: Include stdio.h for sprintf() and
string.h for memcmp().
* ext/dl/ptr.c: Include ctype.h for isdigit().
git-svn-id: svn+ssh://ci.ruby-lang.org/ruby/trunk@3302 b2dd03c8-39d4-4d8f-98ff-823fe69b080e
* ext/extmk.rb (sysquote): do not need to quote on mswin/bccwin/mingw.
* ext/extm.rb ($mflags): uniq items and remove '-' and '--'.
move options to the lead.
* lib/fileutils.rb (install): model on the real install
command(message).
git-svn-id: svn+ssh://ci.ruby-lang.org/ruby/trunk@3292 b2dd03c8-39d4-4d8f-98ff-823fe69b080e
inline function instead so it can be used as an expression.
* eval.c (EXEC_TAG, THREAD_SAVE_CONTEXT): Consistently call
FLUSH_REGISTER_WINDOWS before calling setjmp(). (I suspect that
every setjmp() implementation should take care of register
windows, though)
git-svn-id: svn+ssh://ci.ruby-lang.org/ruby/trunk@3285 b2dd03c8-39d4-4d8f-98ff-823fe69b080e
matz
nobu
eban
ttate
knu
usa