Date: Tue, 4 Sep 2012 16:28:41 +0100 (BST) From: Anton Shterenlikht <mexas@bristol.ac.uk> To: jbeich@tormail.org, mexas@bristol.ac.uk Cc: freebsd-gecko@freebsd.org Subject: Re: ia64 www/firefox - segfault on startup Message-ID: <201209041528.q84FSfOm005248@mech-cluster241.men.bris.ac.uk> In-Reply-To: <1T8dWC-000FCy-Im@internal.tormail.org>
next in thread | previous in thread | raw e-mail | index | archive | help
From jbeich@tormail.org Mon Sep 3 23:37:19 2012 > > It segfaults straight away on the firefox congratulations > page (or whatever it is called): > > http://eis.bris.ac.uk/~mexas/ia64-exp-ff15-debug-output [picking a random thread] > Thread 13 (Thread 134d81800 (LWP 101632/firefox)): > #0 0x000000012049dbc0 in __error () from /lib/libthr.so.3 > #1 0x0000000120482d00 in pthread_getschedparam () from /lib/libthr.so.3 > #2 0x0000000000000000 in ?? () > Previous frame identical to this frame (corrupt stack?) Can you build libthr and nspr with debug symbols? Even if it doesn't trigger an assert gdb backtrace would be more verbose. $ cd /usr/src/lib/libthr; make all install DEBUG_FLAGS='-g' done $ cd /usr/ports/devel/nspr; make install WITH_DEBUG= Also try disabling pthread_getschedparam in devel/nspr. --=-=-= Content-Type: text/x-patch Content-Disposition: attachment; filename=patch-ia64-disable-pthread_getschedparam --- ../pr/src/pthreads/ptthread.c~ +++ ../pr/src/pthreads/ptthread.c @@ -25,7 +25,7 @@ #include <pthread_np.h> #endif -#ifdef SYMBIAN +#if defined(SYMBIAN) || (defined(FREEBSD) && defined(__ia64)) /* In Open C sched_get_priority_min/max do not work properly, so we undefine * _POSIX_THREAD_PRIORITY_SCHEDULING here. */ --=-=-=-- Ok, about devel/nspr - it fails tests. Here's what I did: # pwd /usr/ports/devel/nspr # cat files/patch-ia64-disable-pthread_getschedparam --- ../pr/src/pthreads/ptthread.c~ +++ ../pr/src/pthreads/ptthread.c @@ -25,7 +25,7 @@ #include <pthread_np.h> #endif -#ifdef SYMBIAN +#if defined(SYMBIAN) || (defined(FREEBSD) && defined(__ia64)) /* In Open C sched_get_priority_min/max do not work properly, so we undefine * _POSIX_THREAD_PRIORITY_SCHEDULING here. */ # Fine so far, right? I build WITH_DEBUG=, the log is: http://seis.bris.ac.uk/~mexas/nspr-build.log By the way, can I check just from the build log that I indeed used WITH_DEBUG= ? Is it the "-UDEBUG -DNDEBUG=1" tokens? I never paid much attention to this port, so I never run the tests before. So I don't know if they always failed on ia64, or it is a new thing. Anyway, the test log: http://seis.bris.ac.uk/~mexas/nspr-test.log There are 2 failures: libfilename FAILED libfilename: PR_GetLibraryFilePathname failed libfilename: FAIL peek FAILED peek: Running the test with local threads peek: PR_Connect failed and then "socket" test appears to hang. YOu can see in the log that I started tests at night, and by the morning, the socket was still there. Here's the ps -axwwdd trace: http://seis.bris.ac.uk/~mexas/nspr-test-hang 93280 6 S+ 0:00.09 | `-- script nspr-test.log make test deinstall reinstall 93281 5 Is+ 0:00.20 | `-- make test deinstall reinstall 94911 5 I+ 0:00.25 | `-- /bin/sh ../../../pr/tests/runtests.sh 95023 5 S+ 0:00.37 | `-- ./socket I reinstalled devel/nspr anyway, and will now try to rebuild freebsd-gecko/www/firefox. I'll report back. Thank you Anton P.S. I wonder why the src and ports folk don't agree on a single debug variable. Right now it's DEBUG_FLAGS='-g' in src and WITH_DEBUG= in ports. Is it for a good reason?
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?201209041528.q84FSfOm005248>