[Bug other/26489] [4.1/4.2 Regression] compilation of c++ fails in eh_alloc.cc on NetBSD
--- Comment #15 from sayle at gcc dot gnu dot org 2006-03-02 02:47 --- Subject: Bug 26489 Author: sayle Date: Thu Mar 2 02:47:40 2006 New Revision: 111633 URL: http://gcc.gnu.org/viewcvs?root=gcc&view=rev&rev=111633 Log: PR other/26489 * gthr-posix.h (__gthrw2): Define to take three parameters, the declared name, the weak reference name, and the typeof name. (__gthrw): Avoid expanding the declared name suffix. (__gthrw3): New Tru64 specific macro to simplify the OSF/1 decls. Modified: trunk/gcc/ChangeLog trunk/gcc/gthr-posix.h -- http://gcc.gnu.org/bugzilla/show_bug.cgi?id=26489
[Bug other/26489] [4.1/4.2 Regression] compilation of c++ fails in eh_alloc.cc on NetBSD
--- Comment #14 from dogcow at babymeat dot com 2006-03-02 02:14 --- Created an attachment (id=10953) --> (http://gcc.gnu.org/bugzilla/attachment.cgi?id=10953&action=view) 4.1.0 branch version of patch.txt. -- dogcow at babymeat dot com changed: What|Removed |Added Attachment #10933|0 |1 is obsolete|| http://gcc.gnu.org/bugzilla/show_bug.cgi?id=26489
[Bug other/26489] [4.1/4.2 Regression] compilation of c++ fails in eh_alloc.cc on NetBSD
-- mmitchel at gcc dot gnu dot org changed: What|Removed |Added Target Milestone|4.1.0 |4.1.1 http://gcc.gnu.org/bugzilla/show_bug.cgi?id=26489
[Bug other/26489] [4.1/4.2 Regression] compilation of c++ fails in eh_alloc.cc on NetBSD
--- Comment #13 from dogcow at babymeat dot com 2006-02-28 06:45 --- results sent to gcc-testresults; here's the summary of the g++ section of the tests. FAIL: g++.dg/init/array15.C execution test XPASS: g++.dg/tree-ssa/pr14814.C scan-tree-dump-times &this 0 XPASS: g++.old-deja/g++.other/init5.C execution test === g++ Summary === # of expected passes11965 # of unexpected failures1 # of unexpected successes 2 # of expected failures 66 # of unsupported tests 86 /aux/obj/gcc41/gcc/testsuite/../g++ version 4.1.0 20060228 (prerelease) Running target unix FAIL: 26_numerics/cmath/c99_classification_macros_c++.cc (test for excess errors) FAIL: 26_numerics/cmath/c99_classification_macros_c.cc (test for excess errors) === libstdc++ Summary === # of expected passes2641 # of unexpected failures2 # of expected failures 10 # of unsupported tests 147 -- http://gcc.gnu.org/bugzilla/show_bug.cgi?id=26489
[Bug other/26489] [4.1/4.2 Regression] compilation of c++ fails in eh_alloc.cc on NetBSD
--- Comment #12 from roger at eyesopen dot com 2006-02-28 03:30 --- Hi moof, the way that you can test this patch is to run "make -k check" from the top-level after bootstrapping the tree. You'll notice that even before my change (with RC1 for example), there'll be several thousand libstdc++ and libgfortran failures. With the newly proposed patch, you should now see only a handful of failures for these testsuites. I hope this helps. -- roger at eyesopen dot com changed: What|Removed |Added Keywords||patch http://gcc.gnu.org/bugzilla/show_bug.cgi?id=26489
[Bug other/26489] [4.1/4.2 Regression] compilation of c++ fails in eh_alloc.cc on NetBSD
--- Comment #11 from roger at eyesopen dot com 2006-02-28 03:23 --- Created an attachment (id=10934) --> (http://gcc.gnu.org/bugzilla/attachment.cgi?id=10934&action=view) mainline patch v2 Here is a revised and slightly more tested version of the proposed patch for mainline. The 4.1 branch patch is identical except s/static/external/ in the definition of __gthrw2. 2006-02-27 Roger Sayle <[EMAIL PROTECTED]> PR middle-end/26489 * gthr-posix.h (__gthrw2): Define to take three parameters, the declared name, the weak reference name, and the typeof name. (__gthrw): Avoid expanding the declared name suffix. (__gthrw3): New Tru64 specific macro to simplify the OSF/1 decls. -- roger at eyesopen dot com changed: What|Removed |Added Attachment #10932|0 |1 is obsolete|| http://gcc.gnu.org/bugzilla/show_bug.cgi?id=26489
[Bug other/26489] [4.1/4.2 Regression] compilation of c++ fails in eh_alloc.cc on NetBSD
--- Comment #10 from dogcow at babymeat dot com 2006-02-28 03:02 --- Created an attachment (id=10933) --> (http://gcc.gnu.org/bugzilla/attachment.cgi?id=10933&action=view) patch as of -r 111507 With the (attached) patch modified to work with -r 111507, it now compiles, links, etc after make clean. I'm not sure how to verify that it actually DTRT, however. :) -- http://gcc.gnu.org/bugzilla/show_bug.cgi?id=26489
[Bug other/26489] [4.1/4.2 Regression] compilation of c++ fails in eh_alloc.cc on NetBSD
--- Comment #9 from roger at eyesopen dot com 2006-02-28 02:07 --- Created an attachment (id=10932) --> (http://gcc.gnu.org/bugzilla/attachment.cgi?id=10932&action=view) patch I think this untested patch might fix things. -- http://gcc.gnu.org/bugzilla/show_bug.cgi?id=26489
[Bug other/26489] [4.1/4.2 Regression] compilation of c++ fails in eh_alloc.cc on NetBSD
--- Comment #8 from pinskia at gcc dot gnu dot org 2006-02-28 02:00 ---
Just to show what I am talking about:
extern __typeof(__libc_thr_once) __gthrw_pthread_once __attribute__
((__weakref__("pthread_once")));
That is wrong as __gthrw_pthread_once should be a weak link to __libc_thr_once
and not pthread_once.
So it was wrong before Roger's patch, just Roger's patch exposed a build
failure instead.
--
http://gcc.gnu.org/bugzilla/show_bug.cgi?id=26489
[Bug other/26489] [4.1/4.2 Regression] compilation of c++ fails in eh_alloc.cc on NetBSD
--- Comment #7 from pinskia at gcc dot gnu dot org 2006-02-28 01:48 --- (In reply to comment #5) > So what is happening before, name was not being translated to __libc_thr_once > but only to pthread_once so in fact it looks like it was even wrong in rev > 110280 but might not had noticed it. It was broken in 110280 as obvious by the diff as it was a weak link to pthread_* which is wrong. -- http://gcc.gnu.org/bugzilla/show_bug.cgi?id=26489
[Bug other/26489] [4.1/4.2 Regression] compilation of c++ fails in eh_alloc.cc on NetBSD
--- Comment #6 from dogcow at babymeat dot com 2006-02-28 01:45 --- Created an attachment (id=10931) --> (http://gcc.gnu.org/bugzilla/attachment.cgi?id=10931&action=view) diff of eh_alloc with old and new gthr-posix.h This should be pretty obvious, but it elucidates how the combination of the change from __gthrw to __gthrw2 and the #define pthread_mutexBLAH interact. -- http://gcc.gnu.org/bugzilla/show_bug.cgi?id=26489
[Bug other/26489] [4.1/4.2 Regression] compilation of c++ fails in eh_alloc.cc on NetBSD
--- Comment #5 from pinskia at gcc dot gnu dot org 2006-02-28 01:34 --- I see the bug now: #define pthread_once __libc_thr_once #define pthread_self __libc_thr_self #define pthread_exit __libc_thr_exit #define pthread_setcancelstate __libc_thr_setcancelstate So what is happening before, name was not being translated to __libc_thr_once but only to pthread_once so in fact it looks like it was even wrong in rev 110280 but might not had noticed it. -- http://gcc.gnu.org/bugzilla/show_bug.cgi?id=26489
[Bug other/26489] [4.1/4.2 Regression] compilation of c++ fails in eh_alloc.cc on NetBSD
--- Comment #4 from dogcow at babymeat dot com 2006-02-28 01:29 --- Created an attachment (id=10930) --> (http://gcc.gnu.org/bugzilla/attachment.cgi?id=10930&action=view) -g3 -save-temps'ified eh_alloc.cc -- http://gcc.gnu.org/bugzilla/show_bug.cgi?id=26489
[Bug other/26489] [4.1/4.2 Regression] compilation of c++ fails in eh_alloc.cc on NetBSD
--- Comment #3 from pinskia at gcc dot gnu dot org 2006-02-28 01:18 --- Can you add -g3 -save-temps to the end of the command line and attach the preprocessed source? The -g3 is to also get all the #define's. -- http://gcc.gnu.org/bugzilla/show_bug.cgi?id=26489
