Re: x-www-browser for ultra5 ultra60
On 2009-04-14, Tom spot Callaway tcall...@redhat.com wrote: On 04/13/2009 06:52 PM, Howard Eisenberger wrote: However, apt-get install libnss3-1d also upgrades libsqlite3, and when I started iceweasel, I get a Bus error before it finished loading. It seems to work all right for me with the new nss plus the old libsqlite3. There was an upstream bug in sqlite 3.6.12 that was causing SIGBUS on sparc, there is a patch here: https://bugzilla.redhat.com/show_bug.cgi?id=494906 Thanks. Sorry for the late response. Regards, Howard E. -- To UNSUBSCRIBE, email to debian-sparc-requ...@lists.debian.org with a subject of unsubscribe. Trouble? Contact listmas...@lists.debian.org
Re: x-www-browser for ultra5 ultra60
On 04/13/2009 06:52 PM, Howard Eisenberger wrote: On 2009-04-13, Julien Cristau jcris...@debian.org wrote: It works! With the new libnssd3.so.1d, iceweasel 3.0.7 does not crash on the problem sites on my Ultra 5. The bug was fixed using a different patch in nss 3.12.2.with.ckbi.1.73-2. Thanks. I didn't realize that. However, apt-get install libnss3-1d also upgrades libsqlite3, and when I started iceweasel, I get a Bus error before it finished loading. It seems to work all right for me with the new nss plus the old libsqlite3. There was an upstream bug in sqlite 3.6.12 that was causing SIGBUS on sparc, there is a patch here: https://bugzilla.redhat.com/show_bug.cgi?id=494906 ~spot -- To UNSUBSCRIBE, email to debian-sparc-requ...@lists.debian.org with a subject of unsubscribe. Trouble? Contact listmas...@lists.debian.org
Re: x-www-browser for ultra5 ultra60
On 2009-03-26, Howard Eisenberger wrote: On 2009-03-25, Julien Cristau wrote: It'd be nice if someone could try the attached patch on sparc and see if they can reproduce the browser crash. It works! With the new libnssd3.so.1d, iceweasel 3.0.7 does not crash on the problem sites on my Ultra 5. Has anyone else tried this patch? I have been using it now for almost three weeks without a single crash. Howard E. -- To UNSUBSCRIBE, email to debian-sparc-requ...@lists.debian.org with a subject of unsubscribe. Trouble? Contact listmas...@lists.debian.org
Re: x-www-browser for ultra5 ultra60
On Mon, 2009-04-13 at 12:17 -0400, Howard Eisenberger wrote: On 2009-03-26, Howard Eisenberger wrote: On 2009-03-25, Julien Cristau wrote: It'd be nice if someone could try the attached patch on sparc and see if they can reproduce the browser crash. It works! With the new libnssd3.so.1d, iceweasel 3.0.7 does not crash on the problem sites on my Ultra 5. Has anyone else tried this patch? I have been using it now for almost three weeks without a single crash. The bug was fixed using a different patch in nss 3.12.2.with.ckbi.1.73-2. Cheers, Julien -- To UNSUBSCRIBE, email to debian-sparc-requ...@lists.debian.org with a subject of unsubscribe. Trouble? Contact listmas...@lists.debian.org
Re: x-www-browser for ultra5 ultra60
On 2009-04-13, Julien Cristau jcris...@debian.org wrote: It works! With the new libnssd3.so.1d, iceweasel 3.0.7 does not crash on the problem sites on my Ultra 5. The bug was fixed using a different patch in nss 3.12.2.with.ckbi.1.73-2. Thanks. I didn't realize that. However, apt-get install libnss3-1d also upgrades libsqlite3, and when I started iceweasel, I get a Bus error before it finished loading. It seems to work all right for me with the new nss plus the old libsqlite3. I do have the core dump from the crash and did a trace back to libmozjs if anyone wants to look at it. Regards, Howard E. -- To UNSUBSCRIBE, email to debian-sparc-requ...@lists.debian.org with a subject of unsubscribe. Trouble? Contact listmas...@lists.debian.org
Re: x-www-browser for ultra5 ultra60
On Wed, 2009-03-25 at 21:20 +0100, Mike Hommey wrote: On Wed, Mar 25, 2009 at 09:14:56AM +0100, Julien Cristau wrote: On Wed, 2009-03-25 at 07:47 +0100, Julien Cristau wrote: I've spent some time looking at this, and I'm a bit worried about PKIX_PL_Object_Alloc. Specifically, sizeof(PKIX_PL_Object) seems to be 28 on 32bit, and __alignof__(PKIX_PL_Object) is 4. PKIX_PL_Object_Alloc goes to allocate some space for one PKIX_PL_Object + the size it was asked for, and then goes and returns object + 1. Thus, if PKIX_PL_Malloc gives it a 8 byte aligned pointer, PKIX_PL_Object_Alloc will return an unaligned address to its caller. PKIX_PL_OcspResponse's size is 56, and it has to be 8 byte aligned on sparc, so it's possible this is the problem here. It'd be nice if someone could try the attached patch on sparc and see if they can reproduce the browser crash. Wouldn't it be simpler to make PKIX_PL_Object 32 bytes ? That would work too (hopefully not increasing its 40 bytes on 64bit). Not sure which one is simpler. Cheers, Julien -- To UNSUBSCRIBE, email to debian-sparc-requ...@lists.debian.org with a subject of unsubscribe. Trouble? Contact listmas...@lists.debian.org
Re: x-www-browser for ultra5 ultra60
On Wed, 2009-03-25 at 04:58 +, Howard Eisenberger wrote: On 2009-03-25, Julien Cristau wrote: Core was generated by `/usr/lib/iceweasel/firefox-bin -a iceweasel https://signin.ebay.ca/ws/eBayISAPI'. Program terminated with signal 10, Bus error. #0 0xf69bf6bc in der_TimeStringToTime (dst=0x115f574, string=value optimized out, generalized=2) at dertime.c:311 #1 0xf69bf780 in DER_GeneralizedTimeToTime_Util (dst=0x115f574, time=value optimized out) at dertime.c:239 dertime.c:311 seems to be: *dst = PR_ImplodeTime(genTime); PR_ImplodeTime returns an int64. Here dst is 0x115f574, which is not aligned on a double-word boundary. So that'd explain the SIGBUS. The backtrace you get doesn't go further? It'd be nice to know what the DER_GeneralizedTimeToTime_Util caller is... I saved it this time. Not sure how much you need, but here is a bit more. #0 0xf69bf6bc in der_TimeStringToTime (dst=0x115f574, string=value optimized out, generalized=2) at dertime.c:311 #1 0xf69bf780 in DER_GeneralizedTimeToTime_Util (dst=0x115f574, time=value optimized out) at dertime.c:239 #2 0xf6aa4f54 in pkix_pl_OcspResponse_VerifySignature (response=0x115f554, cert=0x1159b74, procParams=0x11437e4, pPassed=0xf2cea084, #pNBIOContext=0xf2cea078, plContext=0x1143150) at pkix_pl_ocspresponse.c:883 #3 0xf6a5b610 in pkix_OcspChecker_Check (checkerObject=0x1148e8c, #cert=0x1159b74, procParams=0x11437e4, pNBIOContext=0xf2cea19c, #pResultCode=0xf2cea190, plContext=0x1143150) at pkix_ocspchecker.c:266 #4 0xf6a72bb4 in pkix_CheckChain (certs=0x115ad44, numCerts=2, #checkers=0x115ad0c, revCheckers=0x115f0e4, removeCheckedExtOIDs=0x115e8b4, #procParams=0x11437e4, pCertCheckedIndex=0x11588f4, pCheckerIndex=0x11588f8, pRevChecking=0x115891c, pReasonCode=0x1158908, #pNBIOContext=0xf2cea284, pFinalSubjPubKey=0xf2cea290, pPolicyTree=0xf2cea28c, pVerifyTree=0x0, plContext=0x1143150) at pkix_validate.c:354 #5 0xf6a76284 in pkix_Build_ValidateEntireChain (state=0x11588d4, anchor=0x11432ec, pNBIOContext=0xf2cea45c, pValResult=0xf2cea488, #verifyNode=0x1159184, plContext=0x1143150) at pkix_build.c:1619 #6 0xf6a7a734 in pkix_BuildForwardDepthFirstSearch #(pNBIOContext=0xf2cea654, state=0x11588d4, pValResult=0xf2cea64c, #plContext=0x1143150) at pkix_build.c:3052 I've spent some time looking at this, and I'm a bit worried about PKIX_PL_Object_Alloc. Specifically, sizeof(PKIX_PL_Object) seems to be 28 on 32bit, and __alignof__(PKIX_PL_Object) is 4. PKIX_PL_Object_Alloc goes to allocate some space for one PKIX_PL_Object + the size it was asked for, and then goes and returns object + 1. Thus, if PKIX_PL_Malloc gives it a 8 byte aligned pointer, PKIX_PL_Object_Alloc will return an unaligned address to its caller. PKIX_PL_OcspResponse's size is 56, and it has to be 8 byte aligned on sparc, so it's possible this is the problem here. Mike, any idea? Cheers, Julien -- To UNSUBSCRIBE, email to debian-sparc-requ...@lists.debian.org with a subject of unsubscribe. Trouble? Contact listmas...@lists.debian.org
Re: x-www-browser for ultra5 ultra60
On Wed, 2009-03-25 at 07:47 +0100, Julien Cristau wrote: I've spent some time looking at this, and I'm a bit worried about PKIX_PL_Object_Alloc. Specifically, sizeof(PKIX_PL_Object) seems to be 28 on 32bit, and __alignof__(PKIX_PL_Object) is 4. PKIX_PL_Object_Alloc goes to allocate some space for one PKIX_PL_Object + the size it was asked for, and then goes and returns object + 1. Thus, if PKIX_PL_Malloc gives it a 8 byte aligned pointer, PKIX_PL_Object_Alloc will return an unaligned address to its caller. PKIX_PL_OcspResponse's size is 56, and it has to be 8 byte aligned on sparc, so it's possible this is the problem here. It'd be nice if someone could try the attached patch on sparc and see if they can reproduce the browser crash. Cheers, Julien diff -u nss-3.12.2.with.ckbi.1.73/debian/changelog nss-3.12.2.with.ckbi.1.73/debian/changelog --- nss-3.12.2.with.ckbi.1.73/debian/changelog +++ nss-3.12.2.with.ckbi.1.73/debian/changelog @@ -1,3 +1,9 @@ +nss (3.12.2.with.ckbi.1.73-2) UNRELEASED; urgency=low + + * Make sure PKIX_PL_Object_Alloc returns an aligned pointer. + + -- Julien Cristau jcris...@debian.org Wed, 25 Mar 2009 08:38:37 +0100 + nss (3.12.2.with.ckbi.1.73-1) unstable; urgency=low * debian/patches/38_kbsd.dpatch: Brown paper bag fix for regression only in patch2: unchanged: --- nss-3.12.2.with.ckbi.1.73.orig/mozilla/security/nss/lib/libpkix/pkix_pl_nss/system/pkix_pl_object.c +++ nss-3.12.2.with.ckbi.1.73/mozilla/security/nss/lib/libpkix/pkix_pl_nss/system/pkix_pl_object.c @@ -561,6 +561,7 @@ { PKIX_PL_Object *object = NULL; pkix_ClassTable_Entry *ctEntry = NULL; +PKIX_UInt32 alloc_size; PKIX_ENTER(OBJECT, PKIX_PL_Object_Alloc); PKIX_NULLCHECK_ONE(pObject); @@ -605,17 +606,20 @@ PORT_Assert(size == ctEntry-typeObjectSize); -/* Allocate space for the object header and the requested size */ +/* Allocate space for the object header and the requested size, + * and make sure that we return an aligned pointer */ +alloc_size = ((sizeof(PKIX_PL_Object) + 7) ~7) + size; + #ifdef PKIX_OBJECT_LEAK_TEST PKIX_CHECK(PKIX_PL_Calloc (1, -((PKIX_UInt32)sizeof (PKIX_PL_Object))+size, +alloc_size, (void **)object, plContext), PKIX_MALLOCFAILED); #else PKIX_CHECK(PKIX_PL_Malloc -(((PKIX_UInt32)sizeof (PKIX_PL_Object))+size, +(alloc_size, (void **)object, plContext), PKIX_MALLOCFAILED); @@ -641,7 +645,7 @@ /* Return a pointer to the user data. Need to offset by object size */ -*pObject = object + 1; +*pObject = (PKIX_PL_Object *)char*)object) + alloc_size - size)); object = NULL; /* Atomically increment object counter */
Re: x-www-browser for ultra5 ultra60
On Wed, Mar 25, 2009 at 09:14:56AM +0100, Julien Cristau wrote: On Wed, 2009-03-25 at 07:47 +0100, Julien Cristau wrote: I've spent some time looking at this, and I'm a bit worried about PKIX_PL_Object_Alloc. Specifically, sizeof(PKIX_PL_Object) seems to be 28 on 32bit, and __alignof__(PKIX_PL_Object) is 4. PKIX_PL_Object_Alloc goes to allocate some space for one PKIX_PL_Object + the size it was asked for, and then goes and returns object + 1. Thus, if PKIX_PL_Malloc gives it a 8 byte aligned pointer, PKIX_PL_Object_Alloc will return an unaligned address to its caller. PKIX_PL_OcspResponse's size is 56, and it has to be 8 byte aligned on sparc, so it's possible this is the problem here. It'd be nice if someone could try the attached patch on sparc and see if they can reproduce the browser crash. Wouldn't it be simpler to make PKIX_PL_Object 32 bytes ? I do hope there is not more of these in the new PKIX code :-/ Mike -- To UNSUBSCRIBE, email to debian-sparc-requ...@lists.debian.org with a subject of unsubscribe. Trouble? Contact listmas...@lists.debian.org
Re: x-www-browser for ultra5 ultra60
On 2009-03-25, Julien Cristau wrote: It'd be nice if someone could try the attached patch on sparc and see if they can reproduce the browser crash. It works! With the new libnssd3.so.1d, iceweasel 3.0.7 does not crash on the problem sites on my Ultra 5. I used dpkg-buildpackage, which didn't finish, but I copied the new lib from the debian/ subdirectory. I'm not sure how to compile the single lib. Regards, Howard E. -- To UNSUBSCRIBE, email to debian-sparc-requ...@lists.debian.org with a subject of unsubscribe. Trouble? Contact listmas...@lists.debian.org
Re: x-www-browser for ultra5 ultra60
On 2009-03-10, Jurij Smakov wrote: The crashes like this are usually due to unaligned memory accesses, which will cause a SEGFAULT/SIGBUS on sparc, but is tolerated on i386/amd64 platforms. I would not be surprised if that's a manifestation of the bug reported previously [0,1], which, as far as I can tell, is still not fixed upstream. I found that the most useful thing to do is to generate a core file by starting the program in the shell where the core file size is set to unlimited using 'ulimit -c unlimited', then trigger the crash. Given the core file and program executable, gdb should be able to tell where the crash occurs (even better if you have the binary unstripped or have the debugging symbols for it - for example, there is an iceweasel-dbg package in Debian, providing debugging symbols for iceweasel). It is usually fairly easy to tell from the backtrace why the unaligned access happened, usually things like raw memory accesses, for example trying to read a 4-byte word from a location which is not aligned on a 4-byte boundary. Even if it's hard to tell immediately what's wrong, at least it will make a useful bug report. [0] http://bugs.debian.org/362170 [1] https://bugzilla.mozilla.org/show_bug.cgi?id=161826 Thanks for your expertise. I see that the thread for second bug report has continued for over 6 years. I didn't even have Sun hardware when it started. As it turns out, I have been using iceweasel 3.0.7 from unstable for a day now and this is the only page I have encountered so far that crashes the browser. I'll continue to use iceweasel and simply use something else for eBay for the time being. m...@ultra5:~$ ulimit -c unlimited m...@ultra5:~$ iceweasel --verbose https://signin.ebay.ca/ws/eBayISAPI.dll?SignIn ICEWEASEL_DSP= APPLICATION_ID=iceweasel CMDLINE_DISPLAY= DISPLAY=:0.0 OPTIONS=-a iceweasel https://signin.ebay.ca/ws/eBayISAPI.dll?SignIn DEBUG=0 DEBUGGER= MOZ_DISABLE_PANGO= MOZ_NO_REMOTE= Running: /usr/lib/iceweasel/firefox-bin -a iceweasel https://signin.ebay.ca/ws/eBayISAPI.dll?SignIn Bus error (core dumped) m...@ultra5:~$ ls -l core.8659 -rw--- 1 ag221 ag221 80748544 2009-03-24 18:41 core.8659 I have never done any debugging so I'll have to read up on it. Given that I am still using kernel 2.6.21 for X11 on this machine and seeing the issues some others here are having just installing the operating system on their hardware, this browser problem does not seem that significant. Regards, Howard E. -- To UNSUBSCRIBE, email to debian-sparc-requ...@lists.debian.org with a subject of unsubscribe. Trouble? Contact listmas...@lists.debian.org
Re: x-www-browser for ultra5 ultra60
On Tue, Mar 24, 2009 at 11:24:25PM +, Howard Eisenberger wrote: On 2009-03-10, Jurij Smakov wrote: The crashes like this are usually due to unaligned memory accesses, which will cause a SEGFAULT/SIGBUS on sparc, but is tolerated on i386/amd64 platforms. I would not be surprised if that's a manifestation of the bug reported previously [0,1], which, as far as I can tell, is still not fixed upstream. I found that the most useful thing to do is to generate a core file by starting the program in the shell where the core file size is set to unlimited using 'ulimit -c unlimited', then trigger the crash. Given the core file and program executable, gdb should be able to tell where the crash occurs (even better if you have the binary unstripped or have the debugging symbols for it - for example, there is an iceweasel-dbg package in Debian, providing debugging symbols for iceweasel). It is usually fairly easy to tell from the backtrace why the unaligned access happened, usually things like raw memory accesses, for example trying to read a 4-byte word from a location which is not aligned on a 4-byte boundary. Even if it's hard to tell immediately what's wrong, at least it will make a useful bug report. [0] http://bugs.debian.org/362170 [1] https://bugzilla.mozilla.org/show_bug.cgi?id=161826 Thanks for your expertise. I see that the thread for second bug report has continued for over 6 years. I didn't even have Sun hardware when it started. As it turns out, I have been using iceweasel 3.0.7 from unstable for a day now and this is the only page I have encountered so far that crashes the browser. I'll continue to use iceweasel and simply use something else for eBay for the time being. m...@ultra5:~$ ulimit -c unlimited m...@ultra5:~$ iceweasel --verbose https://signin.ebay.ca/ws/eBayISAPI.dll?SignIn ICEWEASEL_DSP= APPLICATION_ID=iceweasel CMDLINE_DISPLAY= DISPLAY=:0.0 OPTIONS=-a iceweasel https://signin.ebay.ca/ws/eBayISAPI.dll?SignIn DEBUG=0 DEBUGGER= MOZ_DISABLE_PANGO= MOZ_NO_REMOTE= Running: /usr/lib/iceweasel/firefox-bin -a iceweasel https://signin.ebay.ca/ws/eBayISAPI.dll?SignIn Bus error (core dumped) m...@ultra5:~$ ls -l core.8659 -rw--- 1 ag221 ag221 80748544 2009-03-24 18:41 core.8659 I have never done any debugging so I'll have to read up on it. It might be as easy is: sudo apt-get install iceweasel-dbg gdb gdb /usr/lib/iceweasel/firefox-bin core.8659 (gdb) bt That should (in theory :-) print out the backtrace of the crash, pointing out the location of the crash and the chain of function calls which lead to it. If it displays something meaningful, post it here and we'll try to figure it out. Given that I am still using kernel 2.6.21 for X11 on this machine and seeing the issues some others here are having just installing the operating system on their hardware, this browser problem does not seem that significant. Regards, Howard E. -- To UNSUBSCRIBE, email to debian-sparc-requ...@lists.debian.org with a subject of unsubscribe. Trouble? Contact listmas...@lists.debian.org -- Jurij Smakov ju...@wooyd.org Key: http://www.wooyd.org/pgpkey/ KeyID: C99E03CC -- To UNSUBSCRIBE, email to debian-sparc-requ...@lists.debian.org with a subject of unsubscribe. Trouble? Contact listmas...@lists.debian.org
Re: x-www-browser for ultra5 ultra60
On 2009-03-25, Jurij Smakov wrote: I have never done any debugging so I'll have to read up on it. It might be as easy is: sudo apt-get install iceweasel-dbg gdb gdb /usr/lib/iceweasel/firefox-bin core.8659 (gdb) bt That should (in theory :-) print out the backtrace of the crash, pointing out the location of the crash and the chain of function calls which lead to it. If it displays something meaningful, post it here and we'll try to figure it out. That was easy enough. Thanks for the quick lesson. xulrunner-1.9 was also upgraded, so I tried it again which reproduced the crash. $ gdb /usr/lib/iceweasel/firefox-bin core.9321 Core was generated by `/usr/lib/iceweasel/firefox-bin -a iceweasel https://signin.ebay.ca/ws/eBayISAPI'. Program terminated with signal 10, Bus error. [New process 9340] [New process 9346] [New process 9345] [New process 9341] [New process 9337] [New process 9336] [New process 9335] [New process 9321] #0 0xf69776bc in ?? () from /usr/lib/libnssutil3.so.1d (gdb) bt #0 0xf69776bc in ?? () from /usr/lib/libnssutil3.so.1d #1 0xf69776bc in ?? () from /usr/lib/libnssutil3.so.1d Backtrace stopped: previous frame identical to this frame (corrupt stack?) (gdb) quit Regards, Howard E. -- To UNSUBSCRIBE, email to debian-sparc-requ...@lists.debian.org with a subject of unsubscribe. Trouble? Contact listmas...@lists.debian.org
Re: x-www-browser for ultra5 ultra60
On Wed, 2009-03-25 at 01:44 +, Howard Eisenberger wrote: #0 0xf69776bc in ?? () from /usr/lib/libnssutil3.so.1d (gdb) bt #0 0xf69776bc in ?? () from /usr/lib/libnssutil3.so.1d #1 0xf69776bc in ?? () from /usr/lib/libnssutil3.so.1d Backtrace stopped: previous frame identical to this frame (corrupt stack?) (gdb) quit You'll want to install libnss3-1d-dbg to get something meaningful here. Cheers, Julien -- To UNSUBSCRIBE, email to debian-sparc-requ...@lists.debian.org with a subject of unsubscribe. Trouble? Contact listmas...@lists.debian.org
Re: x-www-browser for ultra5 ultra60
On 2009-03-25, Julien Cristau jcristau wrote: On Wed, 2009-03-25 at 01:44 +, Howard Eisenberger wrote: #0 0xf69776bc in ?? () from /usr/lib/libnssutil3.so.1d (gdb) bt #0 0xf69776bc in ?? () from /usr/lib/libnssutil3.so.1d #1 0xf69776bc in ?? () from /usr/lib/libnssutil3.so.1d Backtrace stopped: previous frame identical to this frame (corrupt stack?) (gdb) quit You'll want to install libnss3-1d-dbg to get something meaningful here. Right you are. Core was generated by `/usr/lib/iceweasel/firefox-bin -a iceweasel https://signin.ebay.ca/ws/eBayISAPI'. Program terminated with signal 10, Bus error. [New process 2771] #0 0xf69bf6bc in der_TimeStringToTime (dst=0x115f574, string=value optimized out, generalized=2) at dertime.c:311 311 dertime.c: No such file or directory. in dertime.c (gdb) bt #0 0xf69bf6bc in der_TimeStringToTime (dst=0x115f574, string=value optimized out, generalized=2) at dertime.c:311 #1 0xf69bf780 in DER_GeneralizedTimeToTime_Util (dst=0x115f574, time=value optimized out) at dertime.c:239 A quick search for dertime.c (first hit). http://forum.soft32.com/linux/Bug-509930-libnss3-1d-crashes-sparc-ftopict473261.html Problem site: https://homebanking.sskm.de/ . Core was generated by `/usr/lib/iceweasel/firefox-bin -a iceweasel https://homebanking.sskm.de/'. Program terminated with signal 10, Bus error. [New process 2889] #0 0xf697f6bc in der_TimeStringToTime (dst=0x1152684, string=value optimized out, generalized=2) at dertime.c:311 311 dertime.c: No such file or directory. in dertime.c (gdb) bt #0 0xf697f6bc in der_TimeStringToTime (dst=0x1152684, string=value optimized out, generalized=2) at dertime.c:311 #1 0xf697f780 in DER_GeneralizedTimeToTime_Util (dst=0x1152684, time=value optimized out) at dertime.c:239 Thanks for your help. Howard E. -- To UNSUBSCRIBE, email to debian-sparc-requ...@lists.debian.org with a subject of unsubscribe. Trouble? Contact listmas...@lists.debian.org
Re: x-www-browser for ultra5 ultra60
On Wed, 2009-03-25 at 03:55 +, Howard Eisenberger wrote: Core was generated by `/usr/lib/iceweasel/firefox-bin -a iceweasel https://signin.ebay.ca/ws/eBayISAPI'. Program terminated with signal 10, Bus error. [New process 2771] #0 0xf69bf6bc in der_TimeStringToTime (dst=0x115f574, string=value optimized out, generalized=2) at dertime.c:311 311 dertime.c: No such file or directory. in dertime.c (gdb) bt #0 0xf69bf6bc in der_TimeStringToTime (dst=0x115f574, string=value optimized out, generalized=2) at dertime.c:311 #1 0xf69bf780 in DER_GeneralizedTimeToTime_Util (dst=0x115f574, time=value optimized out) at dertime.c:239 dertime.c:311 seems to be: *dst = PR_ImplodeTime(genTime); PR_ImplodeTime returns an int64. Here dst is 0x115f574, which is not aligned on a double-word boundary. So that'd explain the SIGBUS. The backtrace you get doesn't go further? It'd be nice to know what the DER_GeneralizedTimeToTime_Util caller is... Cheers, Julien -- To UNSUBSCRIBE, email to debian-sparc-requ...@lists.debian.org with a subject of unsubscribe. Trouble? Contact listmas...@lists.debian.org
Re: x-www-browser for ultra5 ultra60
On 2009-03-25, Julien Cristau wrote: Core was generated by `/usr/lib/iceweasel/firefox-bin -a iceweasel https://signin.ebay.ca/ws/eBayISAPI'. Program terminated with signal 10, Bus error. #0 0xf69bf6bc in der_TimeStringToTime (dst=0x115f574, string=value optimized out, generalized=2) at dertime.c:311 #1 0xf69bf780 in DER_GeneralizedTimeToTime_Util (dst=0x115f574, time=value optimized out) at dertime.c:239 dertime.c:311 seems to be: *dst = PR_ImplodeTime(genTime); PR_ImplodeTime returns an int64. Here dst is 0x115f574, which is not aligned on a double-word boundary. So that'd explain the SIGBUS. The backtrace you get doesn't go further? It'd be nice to know what the DER_GeneralizedTimeToTime_Util caller is... I saved it this time. Not sure how much you need, but here is a bit more. #0 0xf69bf6bc in der_TimeStringToTime (dst=0x115f574, string=value optimized out, generalized=2) at dertime.c:311 #1 0xf69bf780 in DER_GeneralizedTimeToTime_Util (dst=0x115f574, time=value optimized out) at dertime.c:239 #2 0xf6aa4f54 in pkix_pl_OcspResponse_VerifySignature (response=0x115f554, cert=0x1159b74, procParams=0x11437e4, pPassed=0xf2cea084, #pNBIOContext=0xf2cea078, plContext=0x1143150) at pkix_pl_ocspresponse.c:883 #3 0xf6a5b610 in pkix_OcspChecker_Check (checkerObject=0x1148e8c, #cert=0x1159b74, procParams=0x11437e4, pNBIOContext=0xf2cea19c, #pResultCode=0xf2cea190, plContext=0x1143150) at pkix_ocspchecker.c:266 #4 0xf6a72bb4 in pkix_CheckChain (certs=0x115ad44, numCerts=2, #checkers=0x115ad0c, revCheckers=0x115f0e4, removeCheckedExtOIDs=0x115e8b4, #procParams=0x11437e4, pCertCheckedIndex=0x11588f4, pCheckerIndex=0x11588f8, pRevChecking=0x115891c, pReasonCode=0x1158908, #pNBIOContext=0xf2cea284, pFinalSubjPubKey=0xf2cea290, pPolicyTree=0xf2cea28c, pVerifyTree=0x0, plContext=0x1143150) at pkix_validate.c:354 #5 0xf6a76284 in pkix_Build_ValidateEntireChain (state=0x11588d4, anchor=0x11432ec, pNBIOContext=0xf2cea45c, pValResult=0xf2cea488, #verifyNode=0x1159184, plContext=0x1143150) at pkix_build.c:1619 #6 0xf6a7a734 in pkix_BuildForwardDepthFirstSearch #(pNBIOContext=0xf2cea654, state=0x11588d4, pValResult=0xf2cea64c, #plContext=0x1143150) at pkix_build.c:3052 Regards, Howard E. -- To UNSUBSCRIBE, email to debian-sparc-requ...@lists.debian.org with a subject of unsubscribe. Trouble? Contact listmas...@lists.debian.org
Re: x-www-browser for ultra5 ultra60
On 10 Mar 2009, at 22:21, Jurij Smakov wrote: On Tue, Mar 10, 2009 at 04:58:18AM +, Howard Eisenberger wrote: Thanks for responding. I just tried Epiphany (the browser, not the game) briefly and it also crashed on a problem page (ebay.ca), but I will investigate further. The crashes like this are usually due to unaligned memory accesses, which will cause a SEGFAULT/SIGBUS on sparc, but is tolerated on i386/amd64 platforms. I would not be surprised if that's a manifestation of the bug reported previously [0,1], which, as far as I can tell, is still not fixed upstream. I found that the most useful thing to do is to generate a core file by starting the program in the shell where the core file size is set to unlimited using 'ulimit -c unlimited', then trigger the crash. Given the core file and program executable, gdb should be able to tell where the crash occurs (even better if you have the binary unstripped or have the debugging symbols for it - for example, there is an iceweasel-dbg package in Debian, providing debugging symbols for iceweasel). It is usually fairly easy to tell from the backtrace why the unaligned access happened, usually things like raw memory accesses, for example trying to read a 4-byte word from a location which is not aligned on a 4-byte boundary. Even if it's hard to tell immediately what's wrong, at least it will make a useful bug report. I think the relevant bug(s) for iceweasel have already been reported. It might be worthwhile submitting a bug report for epiphany. I am not a programmer, so won't be able to make much of the core file myself, but I'll have a go at doing this whenever I get a chance. gavin, [0] http://bugs.debian.org/362170 [1] https://bugzilla.mozilla.org/show_bug.cgi?id=161826 -- The thing about words is that meanings can twist just like a snake, and if you want to find snakes look for them behind words that have changed their meaning. -- Terry Pratchett, Lords and Ladies Gavin Duley g...@snoopy.apana.org.au g...@sdf-eu.org WWW: http://gpd.sdf-eu.org/ -- To UNSUBSCRIBE, email to debian-sparc-requ...@lists.debian.org with a subject of unsubscribe. Trouble? Contact listmas...@lists.debian.org
Re: x-www-browser for ultra5 ultra60
On 10 Mar 2009, at 04:58, Howard Eisenberger wrote: On 2009-03-10, gavin duley wrote: At the moment, I am using Epiphany. This is part of the GNOME project. It's not as sophisticated as iceweasel, but it does pretty much everything I need from a browser and works well. http://projects.gnome.org/epiphany/ http://en.wikipedia.org/wiki/Epiphany_(browser) Thanks for responding. I just tried Epiphany (the browser, not the game) briefly and it also crashed on a problem page (ebay.ca), but I will investigate further. That's... odd. I've used Epiphany on my Ultra 5 for quite a while, but apparently I must not have used it for eBay. Epiphany seems to reliably crash every time I click on the link to log in... It's the first major problem I've had with Epiphany so far, though. I tried two other web browsers -- one called netsurf seems to work, but not very well (it's slow and does not have JavaScript support). Another, chimera2, could not handle the eBay login page. So far though, apart from for eBay, Epiphany seems to be the best option for a web browser on the Ultra 5. Out of interest, have you had any problems with X11 on your Ultra 5? I find that unless I use the 2.6.22 kernel, X11 simply doesn't work. I'm running testing (which is probably almost identical to stable about now...) and have had no luck with the 2.6.24 or 2.6.26 kernels. http://lists.humbug.org.au/archives/general/2008-August/027088.html As a matter of fact, I am also using 2.6.22 with testing on this Ultra 5. With 2.6.25 X11 will not run. Okay, thanks. I was wondering if this problem had been fixed by now... gavin, -- As the poet said, Only God can make a tree -- probably because it's so hard to figure out how to get the bark on. -- Woody Allen Gavin Duley g...@snoopy.apana.org.au g...@sdf-eu.org WWW: http://gpd.sdf-eu.org/ -- To UNSUBSCRIBE, email to debian-sparc-requ...@lists.debian.org with a subject of unsubscribe. Trouble? Contact listmas...@lists.debian.org
Re: x-www-browser for ultra5 ultra60
gavin duley a écrit : On 10 Mar 2009, at 04:58, Howard Eisenberger wrote: On 2009-03-10, gavin duley wrote: At the moment, I am using Epiphany. This is part of the GNOME project. It's not as sophisticated as iceweasel, but it does pretty much everything I need from a browser and works well. http://projects.gnome.org/epiphany/ http://en.wikipedia.org/wiki/Epiphany_(browser) Thanks for responding. I just tried Epiphany (the browser, not the game) briefly and it also crashed on a problem page (ebay.ca), but I will investigate further. That's... odd. I've used Epiphany on my Ultra 5 for quite a while, but apparently I must not have used it for eBay. Epiphany seems to reliably crash every time I click on the link to log in... It's the first major problem I've had with Epiphany so far, though. I tried two other web browsers -- one called netsurf seems to work, but not very well (it's slow and does not have JavaScript support). Another, chimera2, could not handle the eBay login page. So far though, apart from for eBay, Epiphany seems to be the best option for a web browser on the Ultra 5. On my U60 and my U2, I have rebuilt seamonkey from sources. 1.1.14 runs fine. 2.0a3 craches very early with a SIGBUS. Regards, JKB -- To UNSUBSCRIBE, email to debian-sparc-requ...@lists.debian.org with a subject of unsubscribe. Trouble? Contact listmas...@lists.debian.org
Re: x-www-browser for ultra5 ultra60
On Tue, Mar 10, 2009 at 04:58:18AM +, Howard Eisenberger wrote: On 2009-03-10, gavin duley wrote: Howard Eisenberger wrote: This is a general question. I have been unable to run iceweasel on my ultra5 or ultra60 since it went from 2.0 to 3.0. It crashes constantly with Bus error. I've had a similar problem with iceweasel on my Ultra 5, apparently due to a futex bug: http://lists.debian.org/debian-sparc/2007/07/msg00130.html I have been using iceape (currently version 1.1.14). Now I see that iceape has been dropped from Lenny. I tried midori briefly on unstable, but it also crashed. Is anyone having more success in this regard on similar hardware? At the moment, I am using Epiphany. This is part of the GNOME project. It's not as sophisticated as iceweasel, but it does pretty much everything I need from a browser and works well. http://projects.gnome.org/epiphany/ http://en.wikipedia.org/wiki/Epiphany_(browser) Thanks for responding. I just tried Epiphany (the browser, not the game) briefly and it also crashed on a problem page (ebay.ca), but I will investigate further. The crashes like this are usually due to unaligned memory accesses, which will cause a SEGFAULT/SIGBUS on sparc, but is tolerated on i386/amd64 platforms. I would not be surprised if that's a manifestation of the bug reported previously [0,1], which, as far as I can tell, is still not fixed upstream. I found that the most useful thing to do is to generate a core file by starting the program in the shell where the core file size is set to unlimited using 'ulimit -c unlimited', then trigger the crash. Given the core file and program executable, gdb should be able to tell where the crash occurs (even better if you have the binary unstripped or have the debugging symbols for it - for example, there is an iceweasel-dbg package in Debian, providing debugging symbols for iceweasel). It is usually fairly easy to tell from the backtrace why the unaligned access happened, usually things like raw memory accesses, for example trying to read a 4-byte word from a location which is not aligned on a 4-byte boundary. Even if it's hard to tell immediately what's wrong, at least it will make a useful bug report. [0] http://bugs.debian.org/362170 [1] https://bugzilla.mozilla.org/show_bug.cgi?id=161826 Best regards, -- Jurij Smakov ju...@wooyd.org Key: http://www.wooyd.org/pgpkey/ KeyID: C99E03CC -- To UNSUBSCRIBE, email to debian-sparc-requ...@lists.debian.org with a subject of unsubscribe. Trouble? Contact listmas...@lists.debian.org
Re: x-www-browser for ultra5 ultra60
Howard Eisenberger wrote: This is a general question. I have been unable to run iceweasel on my ultra5 or ultra60 since it went from 2.0 to 3.0. It crashes constantly with Bus error. I've had a similar problem with iceweasel on my Ultra 5, apparently due to a futex bug: http://lists.debian.org/debian-sparc/2007/07/msg00130.html I have been using iceape (currently version 1.1.14). Now I see that iceape has been dropped from Lenny. I tried midori briefly on unstable, but it also crashed. Is anyone having more success in this regard on similar hardware? At the moment, I am using Epiphany. This is part of the GNOME project. It's not as sophisticated as iceweasel, but it does pretty much everything I need from a browser and works well. http://projects.gnome.org/epiphany/ http://en.wikipedia.org/wiki/Epiphany_(browser) Out of interest, have you had any problems with X11 on your Ultra 5? I find that unless I use the 2.6.22 kernel, X11 simply doesn't work. I'm running testing (which is probably almost identical to stable about now...) and have had no luck with the 2.6.24 or 2.6.26 kernels. http://lists.humbug.org.au/archives/general/2008-August/027088.html gavin, -- Gavin Duley g...@snoopy.apana.org.au g...@sdf-eu.org WWW: http://gpd.sdf-eu.org/ -- To UNSUBSCRIBE, email to debian-sparc-requ...@lists.debian.org with a subject of unsubscribe. Trouble? Contact listmas...@lists.debian.org
Re: x-www-browser for ultra5 ultra60
On 2009-03-10, gavin duley wrote: Howard Eisenberger wrote: This is a general question. I have been unable to run iceweasel on my ultra5 or ultra60 since it went from 2.0 to 3.0. It crashes constantly with Bus error. I've had a similar problem with iceweasel on my Ultra 5, apparently due to a futex bug: http://lists.debian.org/debian-sparc/2007/07/msg00130.html I have been using iceape (currently version 1.1.14). Now I see that iceape has been dropped from Lenny. I tried midori briefly on unstable, but it also crashed. Is anyone having more success in this regard on similar hardware? At the moment, I am using Epiphany. This is part of the GNOME project. It's not as sophisticated as iceweasel, but it does pretty much everything I need from a browser and works well. http://projects.gnome.org/epiphany/ http://en.wikipedia.org/wiki/Epiphany_(browser) Thanks for responding. I just tried Epiphany (the browser, not the game) briefly and it also crashed on a problem page (ebay.ca), but I will investigate further. Out of interest, have you had any problems with X11 on your Ultra 5? I find that unless I use the 2.6.22 kernel, X11 simply doesn't work. I'm running testing (which is probably almost identical to stable about now...) and have had no luck with the 2.6.24 or 2.6.26 kernels. http://lists.humbug.org.au/archives/general/2008-August/027088.html As a matter of fact, I am also using 2.6.22 with testing on this Ultra 5. With 2.6.25 X11 will not run. Regards, Howard E. Ottawa -- To UNSUBSCRIBE, email to debian-sparc-requ...@lists.debian.org with a subject of unsubscribe. Trouble? Contact listmas...@lists.debian.org