use NEWLINE flag instead.
* ext/socket/socket.c (sock_gethostbyname): returns host if
ai_canonname is NULL. (ruby-bugs PR#1243)
* parse.y (block_append): update nd_end for "real" head node.
[ruby-list:39058]
* marshal.c (w_class): should not dump singleton class.
[ruby-dev:22631]
git-svn-id: svn+ssh://ci.ruby-lang.org/ruby/trunk@5535 b2dd03c8-39d4-4d8f-98ff-823fe69b080e
process to get the base address for the RSE backing store, so
hardcode it for the moment.
[submitted by: Marcel Moolenaar <marcel@FreeBSD.org>]
git-svn-id: svn+ssh://ci.ruby-lang.org/ruby/trunk@5039 b2dd03c8-39d4-4d8f-98ff-823fe69b080e
pthread is used.
* marshal.c (w_extended): singleton methods should not be checked
when dumping via marshal_dump() or _dump(). [ruby-talk:85909]
* file.c (getcwdofdrv): avoid using getcwd() directly, use
my_getcwd() instead.
* merged NeXT, OpenStep, Rhapsody ports patch from Eric Sunshine
<sunshine@sunshineco.com>. [ruby-core:01596]
* marshal.c (w_object): LINK check earlier than anything else,
i.e. do not dump TYPE_IVAR for already dumped objects.
(ruby-bugs PR#1220)
* eval.c (rb_eval): call "inherited" only when a new class is
generated; not on reopening.
* eval.c (eval): prepend error position in evaluating string to
* configure.in: revived NextStep, OpenStep, and Rhapsody ports which
had become unbuildable; enhanced --enable-fat-binary option so that
it accepts a list of desired architectures (rather than assuming a
fixed list), or defaults to a platform-appropriate list if user does
not provide an explicit list; made the default list of architectures
for MAB (fat binary) more comprehensive; now uses -fno-common even
when building the interpreter (in addition to using it for
extensions), thus allowing the interpreter to be embedded into a
plugin module of an external project (in addition to allowing
embedding directly into an application); added checks for
<netinet/in_systm.h> (needed by `socket' extension) and getcwd(); now
ensures that -I/usr/local/include is employed when extensions'
extconf.rb scripts invoke have_header() since extension checks on
NextStep and OpenStep will fail without it if the desired resource
resides in the /usr/local tree; fixed formatting of --help message.
* Makefile.in: $(LIBRUBY_A) rule now deletes the archive before
invoking $(AR) since `ar' on Apple/NeXT can not "update" MAB archives
(see configure's --enable-fat-binary option); added rule for new
missing/getcwd.c.
* defines.h: fixed endian handling during MAB build (see configure's
--enable-fat-binary option) to ensure that all portions of the
project see the correct WORDS_BIGENDIAN value (some extension modules
were getting the wrong endian setting); added missing constants
GETPGRP_VOID, WNOHANG, WUNTRACED, X_OK, and type pid_t for NextStep
and OpenStep; removed unnecessary and problematic HAVE_SYS_WAIT_H
define in NeXT section.
* dir.c: do not allow NAMLEN() macro to trust dirent::d_namlen on
NextStep since, on some installations, this value always resolves
uselessly to zero.
* dln.c: added error reporting to NextStep extension loader since the
previous behavior of failing silently was not useful; now ensures
that NSLINKMODULE_OPTION_BINDNOW compatibility constant is defined
for OpenStep and Rhapsody; no longer includes <mach-o/dyld.h> twice
on Rhapsody since this header lacks multiple-include protection,
which resulted in "redefinition" compilation errors.
* main.c: also create hard reference to objc_msgSend() on NeXT
platforms (in addition to Apple platforms).
* lib/mkmf.rb: now exports XCFLAGS from configure script to extension
makefiles so that extensions can be built MAB (see configure's
--enable-fat-binary option); also utilize XCFLAGS in cc_command()
(but not cpp_command() because MAB flags are incompatible with
direct invocation of `cpp').
* ext/curses/extconf.rb: now additionally checks for presence of these
curses functions which are not present on NextStep or Openstep:
bkgd(), bkgdset(), color(), curs(), getbkgd(), init(), scrl(), set(),
setscrreg(), wattroff(), wattron(), wattrset(), wbkgd(), wbkgdset(),
wscrl(), wsetscrreg()
* ext/curses/curses.c: added appropriate #ifdef's for additional set of
curses functions now checked by extconf.rb; fixed curses_bkgd() and
window_bkgd() to correctly return boolean result rather than numeric
result; fixed window_getbkgd() to correctly signal an error by
returning nil rather than -1.
* ext/etc/etc.c: setup_passwd() and setup_group() now check for null
pointers before invoking rb_tainted_str_new2() upon fields extracted
from `struct passwd' and `struct group' since null pointers in some
fields are common on NextStep/OpenStep (especially so for the
`pw_comment' field) and rb_tainted_str_new2() throws an exception
when it receives a null pointer.
* ext/pty/pty.c: include "util.h" for strdup()/ruby_strdup() for
platforms such as NextStep and OpenStep which lack strdup().
* ext/socket/getaddrinfo.c: cast first argument of getservbyname(),
gethostbyaddr(), and gethostbyname() from (const char*) to non-const
(char*) for older platforms such as NextStep and OpenStep.
* ext/socket/socket.c: include "util.h" for strdup()/ruby_strdup() for
platforms such as NextStep and OpenStep which lack strdup(); include
<netinet/in_systm.h> if present for NextStep and OpenStep; cast first
argument of gethostbyaddr() and getservbyname() from (const char*) to
non-const (char*) for older platforms.
* ext/syslog/syslog.c: include "util.h" for strdup()/ruby_strdup() for
platforms such as NextStep and OpenStep which lack strdup().
git-svn-id: svn+ssh://ci.ruby-lang.org/ruby/trunk@5002 b2dd03c8-39d4-4d8f-98ff-823fe69b080e
race condition. [ruby-dev:21633]
* gc.c (gc_sweep): do not update malloc_limit when malloc_increase
is smaller than malloc_limit.
* ext/socket/socket.c (make_hostent): h_aliases may be NULL.
(ruby-bugs PR#1195)
* ext/socket/socket.c (sock_s_gethostbyaddr): ditto.
git-svn-id: svn+ssh://ci.ruby-lang.org/ruby/trunk@4808 b2dd03c8-39d4-4d8f-98ff-823fe69b080e
* gc.c (add_heap): should not clear heaps slot even if realloc()
failed.
git-svn-id: svn+ssh://ci.ruby-lang.org/ruby/trunk@4639 b2dd03c8-39d4-4d8f-98ff-823fe69b080e
[ruby-core:01408]
* eval.c (Init_Thread): Continuation#[] added. [ruby-talk:79028]
* parse.y (mlhs_node): should allow "::Foo" (colon3) as lhs.
* parse.y (lhs): ditto.
* parse.y (yylex): should return tCOLON3 right after kCLASS.
[ruby-talk:78918]
* error.c (exc_initialize): was converting argument to string too
eagerly. Only check was needed. [ruby-talk:78958]
git-svn-id: svn+ssh://ci.ruby-lang.org/ruby/trunk@4390 b2dd03c8-39d4-4d8f-98ff-823fe69b080e
Linux/ia64. This makes gc.c compile but miniruby coredumps for
the moment.
git-svn-id: svn+ssh://ci.ruby-lang.org/ruby/trunk@4345 b2dd03c8-39d4-4d8f-98ff-823fe69b080e
"self", making it default to "int".
* ext/syck/rubyext.c (syck_emitter_simple_write): ditto.
* gc.c (rb_gc): should mark backing store region on IA64.
git-svn-id: svn+ssh://ci.ruby-lang.org/ruby/trunk@4266 b2dd03c8-39d4-4d8f-98ff-823fe69b080e
magic number when optimizer turned on, regardless of rlimit
values.
git-svn-id: svn+ssh://ci.ruby-lang.org/ruby/trunk@4247 b2dd03c8-39d4-4d8f-98ff-823fe69b080e
warned for "void value expression". [ruby-talk:72989]
* gc.c (add_final): should determine type by respond_to?
* gc.c (define_final): ditto.
* io.c (rb_io_ctl): should not depend on respond_to?
* range.c (range_step): rb_check_string_type().
* process.c (proc_setgroups): new functions.
git-svn-id: svn+ssh://ci.ruby-lang.org/ruby/trunk@3918 b2dd03c8-39d4-4d8f-98ff-823fe69b080e
[ruby-dev:20272]
* marshal.c: try to make ArgumentError and TypeError consistent.
[ruby-core:01068]
git-svn-id: svn+ssh://ci.ruby-lang.org/ruby/trunk@3853 b2dd03c8-39d4-4d8f-98ff-823fe69b080e
subclasses of SystemCallError. in addition, if initialize is
called for SystenCallError instance, its class be changed.
[ruby-dev:20257]
* gc.c (run_final): to protect thread context switch, finalizers
are wrapped in DEFER_INTS/ENABLE_INTS.
git-svn-id: svn+ssh://ci.ruby-lang.org/ruby/trunk@3839 b2dd03c8-39d4-4d8f-98ff-823fe69b080e
addresses. C variables may not hold valid reference to Ruby
objects. [ruby-core:00975]
git-svn-id: svn+ssh://ci.ruby-lang.org/ruby/trunk@3705 b2dd03c8-39d4-4d8f-98ff-823fe69b080e
* parse.y (primary): "self[n]=x" can be legal even when "[]=" is
private. changes submitted in [ruby-talk:63982]
* parse.y (aryset): ditto.
* parse.y (attrset): "self.foo=x" can be legal even when "foo="
is private.
* eval.c (is_defined): private "[]=" and "foo=" support.
* eval.c (rb_eval): ditto.
* eval.c (assign): ditto.
* eval.c (rb_eval): "foo=" should not always be public.
git-svn-id: svn+ssh://ci.ruby-lang.org/ruby/trunk@3509 b2dd03c8-39d4-4d8f-98ff-823fe69b080e
matz
dave
knu
nagai
eban
nobu
usa
uema2
michal