[opensuse-factory] gcc miscompilation on Factory i586?
Hi, I've got a real weird issue by trying to update mozilla-nss for Factory what I can only interpret as compiler issue or very weird coding issue. mozilla-nss is running a comprehensive test suite during build which currently only seems to fail on openSUSE_Factory i586. It apparently works on x86_64, all older distributions and also in Factory_staging. https://build.opensuse.org/package/show?package=mozilla-nss&project=mozilla%3AFactory In the failing buildlog one can see that it's crashing on a tool called signtool with free() on an invalid pointer. (search for "- FAILED"). When I build locally I also get those cores dumped and looked into two of them which show the following stack: (gdb) bt #0 0xb7be39d6 in raise () from /lib/libc.so.6 #1 0xb7be52d8 in abort () from /lib/libc.so.6 #2 0xb7c1fa25 in ?? () from /lib/libc.so.6 #3 0xb7c259c5 in ?? () from /lib/libc.so.6 #4 0xb7c2727b in free () from /lib/libc.so.6 #5 0xb7d45ab7 in PR_Free () from /usr/lib/libnspr4.so #6 0xb7d89f41 in PORT_Free_Util (ptr=0x0) at secport.c:152 #7 0x0805b146 in jar_extract_mf (jar=0x8091f90, format=<value optimized out>, fp=0x8092048, ext=0x806a210 "sf") at jarfile.c:737 #8 0x0805b5b2 in jar_extract_manifests (fp=<value optimized out>, format=<value optimized out>, jar=<value optimized out>) at jarfile.c:611 #9 JAR_pass_archive (jar=0x8091f90, format=jarArchZip, filename=0x80795c8 "nojs.jar", url=0x806c5f1 "some-url") at jarfile.c:134 #10 0x08057c32 in VerifyJar (filename=0x80795c8 "nojs.jar") at verify.c:74 #11 0x0804e7fe in main (argc=9, argv=0xbfa54424) at signtool.c:1007 Looking at secport.c:152 I find: void PORT_Free(void *ptr) { if (ptr) { PR_Free(ptr); } } I'm really confused how the above stack can happen at all if ptr=0x0. That's the reason why I'm actually thinking about some weird toolchain thing. That's a _bit_ important to me as I need to update NSS for different reasons and I don't want to disable the testsuite. Any pointers for me on that? Thanks, Wolfgang -- To unsubscribe, e-mail: opensuse-factory+unsubscribe@opensuse.org For additional commands, e-mail: opensuse-factory+help@opensuse.org
On Tue, Apr 21, 2009 at 04:24:40PM +0200, Wolfgang Rosenauer wrote:
Hi,
I've got a real weird issue by trying to update mozilla-nss for Factory what I can only interpret as compiler issue or very weird coding issue.
mozilla-nss is running a comprehensive test suite during build which currently only seems to fail on openSUSE_Factory i586.
It apparently works on x86_64, all older distributions and also in Factory_staging.
https://build.opensuse.org/package/show?package=mozilla-nss&project=mozilla%3AFactory
In the failing buildlog one can see that it's crashing on a tool called signtool with free() on an invalid pointer. (search for "- FAILED").
When I build locally I also get those cores dumped and looked into two of them which show the following stack:
(gdb) bt #0 0xb7be39d6 in raise () from /lib/libc.so.6 #1 0xb7be52d8 in abort () from /lib/libc.so.6 #2 0xb7c1fa25 in ?? () from /lib/libc.so.6 #3 0xb7c259c5 in ?? () from /lib/libc.so.6 #4 0xb7c2727b in free () from /lib/libc.so.6 #5 0xb7d45ab7 in PR_Free () from /usr/lib/libnspr4.so #6 0xb7d89f41 in PORT_Free_Util (ptr=0x0) at secport.c:152 #7 0x0805b146 in jar_extract_mf (jar=0x8091f90, format=<value optimized out>, fp=0x8092048, ext=0x806a210 "sf") at jarfile.c:737 #8 0x0805b5b2 in jar_extract_manifests (fp=<value optimized out>, format=<value optimized out>, jar=<value optimized out>) at jarfile.c:611 #9 JAR_pass_archive (jar=0x8091f90, format=jarArchZip, filename=0x80795c8 "nojs.jar", url=0x806c5f1 "some-url") at jarfile.c:134 #10 0x08057c32 in VerifyJar (filename=0x80795c8 "nojs.jar") at verify.c:74 #11 0x0804e7fe in main (argc=9, argv=0xbfa54424) at signtool.c:1007
Looking at secport.c:152 I find: void PORT_Free(void *ptr) { if (ptr) { PR_Free(ptr); } }
I'm really confused how the above stack can happen at all if ptr=0x0.
That's the reason why I'm actually thinking about some weird toolchain thing. That's a _bit_ important to me as I need to update NSS for different reasons and I don't want to disable the testsuite.
Any pointers for me on that?
free (NULL); is a good operation and will not abort(). ptr is not NULL at this point in time I think, the backtrace is just a bit erroneous. However, you corrupted memory earlier somewhere. Can you run this in valgrind? Ciao, Marcus -- To unsubscribe, e-mail: opensuse-factory+unsubscribe@opensuse.org For additional commands, e-mail: opensuse-factory+help@opensuse.org
Marcus Meissner wrote:
However, you corrupted memory earlier somewhere.
Can you run this in valgrind?
We definitely tend to run into risks of memory corruptions sometimes with Mozilla code, on my Factory (i686) system I consistently run into a memory corruption crash (free invalid pointer) of some sort when trying to open print of print preview from my self-compiled builds based on 1.9.1 or later Mozilla code (no matter if Firefox, SeaMonkey, or Thunderbird). I haven't tried builds from OBS yet, but I know the Firefox 3.0 install work, which should be from a Mozilla repo on OBS. The crashes look to be connected to jemalloc somehow, but the same sources compiled on a 11.1 machine (my laptop) work without problems. If there might be a problem with toolchain or compiler, it could affect my case as well for sure. I haven't run things in valgrind, I don't know my way around with it really and have only installed it now so someone might be able to guide me there some time. Robert Kaiser -- To unsubscribe, e-mail: opensuse-factory+unsubscribe@opensuse.org For additional commands, e-mail: opensuse-factory+help@opensuse.org
Marcus Meissner wrote:
On Tue, Apr 21, 2009 at 04:24:40PM +0200, Wolfgang Rosenauer wrote:
I've got a real weird issue by trying to update mozilla-nss for Factory what I can only interpret as compiler issue or very weird coding issue.
As a followup on this, I'm now seeing Firefox 3.5b4 from current Factory crashing at Print Preview and Print, is anyone else seeing this? If yes, is a bug filed? Robert Kaiser -- To unsubscribe, e-mail: opensuse-factory+unsubscribe@opensuse.org For additional commands, e-mail: opensuse-factory+help@opensuse.org
Robert Kaiser schrieb:
Marcus Meissner wrote:
On Tue, Apr 21, 2009 at 04:24:40PM +0200, Wolfgang Rosenauer wrote:
I've got a real weird issue by trying to update mozilla-nss for Factory what I can only interpret as compiler issue or very weird coding issue.
As a followup on this, I'm now seeing Firefox 3.5b4 from current Factory crashing at Print Preview and Print, is anyone else seeing this? If yes, is a bug filed?
I just tried and get the same with Print preview. There is no bug filed, so if you like to... Thanks for the report, Wolfgang -- To unsubscribe, e-mail: opensuse-factory+unsubscribe@opensuse.org For additional commands, e-mail: opensuse-factory+help@opensuse.org
Wolfgang Rosenauer schrieb:
Robert Kaiser schrieb:
Marcus Meissner wrote:
On Tue, Apr 21, 2009 at 04:24:40PM +0200, Wolfgang Rosenauer wrote:
I've got a real weird issue by trying to update mozilla-nss for Factory what I can only interpret as compiler issue or very weird coding issue. As a followup on this, I'm now seeing Firefox 3.5b4 from current Factory crashing at Print Preview and Print, is anyone else seeing this? If yes, is a bug filed?
I just tried and get the same with Print preview. There is no bug filed, so if you like to...
I have to correct myself after looking at the backtrace. A bug is filed at https://bugzilla.novell.com/show_bug.cgi?id=503151 actually. Wolfgang -- To unsubscribe, e-mail: opensuse-factory+unsubscribe@opensuse.org For additional commands, e-mail: opensuse-factory+help@opensuse.org
Wolfgang Rosenauer wrote:
I have to correct myself after looking at the backtrace. A bug is filed at https://bugzilla.novell.com/show_bug.cgi?id=503151 actually.
Thanks for the pointer. I have seen this crash from print preview for a while with all my self-compiled Mozilla products on Factory, having the switch from Firefox 3.0 to 3.5 expose it to others at least shows me it's not just my system, which is good to know. I have added my knowledge about the issue to the bug. I just hope people don't try printing web pages from Factory too much until a solution for that can be found. Robert Kaiser -- To unsubscribe, e-mail: opensuse-factory+unsubscribe@opensuse.org For additional commands, e-mail: opensuse-factory+help@opensuse.org
Wolfgang Rosenauer wrote:
Wolfgang Rosenauer schrieb:
Robert Kaiser schrieb:
Marcus Meissner wrote:
On Tue, Apr 21, 2009 at 04:24:40PM +0200, Wolfgang Rosenauer wrote:
I've got a real weird issue by trying to update mozilla-nss for Factory what I can only interpret as compiler issue or very weird coding issue. As a followup on this, I'm now seeing Firefox 3.5b4 from current Factory crashing at Print Preview and Print, is anyone else seeing this? If yes, is a bug filed? I just tried and get the same with Print preview. There is no bug filed, so if you like to...
I have to correct myself after looking at the backtrace. A bug is filed at https://bugzilla.novell.com/show_bug.cgi?id=503151 actually.
Wolfgang
Print is totally busted at the moment. I have been able to print from kubuntu in a VirtualBox VM, but not natively or across the network from other 11.2 boxes. My latest efforts of recompiling the rpm source for ghostscript library only resulted in no print and the job stays on the queue - no error or access problems. Previously I was seeing a crash in pstoraster. Scan is fine. I've tried a HP Photosmart C4180 and a Lexmark z600, the Lexmark ejects a blank page, otherwise it's the same as for the HP. Regards Sid. -- Sid Boyce ... Hamradio License G3VBV, Licensed Private Pilot Emeritus IBM/Amdahl Mainframes and Sun/Fujitsu Servers Tech Support Specialist, Cricket Coach Microsoft Windows Free Zone - Linux used for all Computing Tasks -- To unsubscribe, e-mail: opensuse-factory+unsubscribe@opensuse.org For additional commands, e-mail: opensuse-factory+help@opensuse.org
Robert Kaiser wrote:
Wolfgang Rosenauer wrote:
I have to correct myself after looking at the backtrace. A bug is filed at https://bugzilla.novell.com/show_bug.cgi?id=503151 actually.
Thanks for the pointer. I have seen this crash from print preview for a while with all my self-compiled Mozilla products on Factory, having the switch from Firefox 3.0 to 3.5 expose it to others at least shows me it's not just my system, which is good to know. I have added my knowledge about the issue to the bug.
I just hope people don't try printing web pages from Factory too much until a solution for that can be found.
Robert Kaiser
More to the point, can you print at all and what printer are you using? Regards Sid. -- Sid Boyce ... Hamradio License G3VBV, Licensed Private Pilot Emeritus IBM/Amdahl Mainframes and Sun/Fujitsu Servers Tech Support Specialist, Cricket Coach Microsoft Windows Free Zone - Linux used for all Computing Tasks -- To unsubscribe, e-mail: opensuse-factory+unsubscribe@opensuse.org For additional commands, e-mail: opensuse-factory+help@opensuse.org
Sid Boyce wrote:
More to the point, can you print at all and what printer are you using?
With Firefox or other Mozilla apps, that's not even the question right now, they already crash before showing any printer selection or other printer dialog, or even print preview - which is the problem we were talking about here. Printing in general is broken in a way that my printer (Canon PIXMA ip4000R, either with Canon-Japan's GPL driver or with Gutenprint) prints only garbage. The problem with Mozilla apps crashing was already existent with my self-built development versions of those when printing from other applications still worked. Right now I'm glad that this printer has network support and I can use it from my laptop that has openSUSE 11.1 installed - both general printing and printing from Mozilla applications work fine there. Robert Kaiser -- To unsubscribe, e-mail: opensuse-factory+unsubscribe@opensuse.org For additional commands, e-mail: opensuse-factory+help@opensuse.org
participants (4)
-
Marcus Meissner -
Robert Kaiser -
Sid Boyce -
Wolfgang Rosenauer