Closed Bug 183907 Opened 22 years ago Closed 22 years ago

AIX 1.2.1 Zilla dumped core on startup

Categories

(SeaMonkey :: General, defect)

Product:
SeaMonkey
Component:
General
Platform:
Other
AIX
Type:
defect
Priority:
Not set
Severity:
normal

Tracking

(Not tracked)

Status:
VERIFIED FIXED

People

(Reporter: ai, Assigned: pkwarren)

References

Details

Attachments

(1 file)

Increases AIX stack from 64k to 128k
777 bytes, patch
roland.mainz
: review+
wtc
: superreview+
Details | Diff | Splinter Review
User-Agent: Mozilla/4.8 [en] (X11; U; AIX 4.3) Build Identifier: Mozilla 1.2.1 for AIX 4.3.3 Mozilla 1.2.1 core dumped being start on AIX 4.3.3 ML10 + _LAST_FIXES_ (libc 4.3.3.88, kernel 4.3.3.86 & ...) With previous libc/kernel/etc, Mozilla worked ok. Reproducible: Always Steps to Reproduce: 1. 2. 3.
Reporter: Do you have any stack-trace from the coredump (if you have dbx installed the output of % /bin/echo "run\nwhere" | ./mozilla -g -d dbx # may be usefull) ?
...skip... warning: Unable to access address 0xf05af88c from core warning: Unable to access address 0xf05af890 from core warning: Unable to access address 0xf05af890 from core warning: no source compiled with -g [using memory image in core] warning: Unable to access address 0xf03488e4 from core Segmentation fault in is_data_there at 0xd020e5fc 0xd020e5fc (is_data_there+0x1c) 9421ffc0 stwu r1,-64(r1) (/usr/ccs/bin/dbx) where is_data_there() at 0xd020e5fc __which_res_state() at 0xd020eacc ho_byname2(??, ??, ??) at 0xd022bbd8 ho_byname2(??, ??, ??) at 0xd0241ef0 gethostbyname2(??, ??) at 0xd020dc34 SOMObjectNewClass() at 0xd12334e8 g_timer_new() at 0xd1c3f4b0 g_timer_new() at 0xd1c3f4f8 g_timer_new() at 0xd1c4162c dowprnt.mbcounter() at 0xd17cfb30 SOMObjectNewClass() at 0xd1221e54 pthread._pthread_body(??) at 0xd00080c8 (/usr/ccs/bin/dbx)
Mhhh, I have two emails with reports about gtimer crashes, too: -- snip -- Segmentation fault in gtkinvisible.g_error at line 1329 in file "" ($t1) could not read "gtkinvisible.c" gtkinvisible.g_error(format = (nil)), line 1329 in "gtkinvisible.c" gtk_set_locale(), line 444 in "gtkmain.c" gtimer.g_error() at 0x21471ab4 gtimer.g_error() at 0x2141b6e0 .() at 0x10003574 .() at 0x10002a68 -- snip --
Summary: core dumped being start → AIX 1.2.1 Zilla dumped core on startup
Have you tried starting mozilla with the Profile Manager? ./mozilla -profileManager Sometimes we see some sort of race condition on startup in AIX Mozilla, and it helps to either create a new profile or start initially with the Profile Manager.
additional info: home directory is mounted via NFS. old profile: ------------------------------------------------------ $ /opt/mozilla/mozilla -profileManager /opt/mozilla/run-mozilla.sh[36]: 31460 Segmentation fault(coredump) Oh no! /opt/mozilla/mozilla-bin just dumped a core file. Do you want to debug this ? You need a lot of memory for this, so watch out ? [y/n] y ...skip... warning: Unable to access address 0xf0448890 from core warning: Unable to access address 0xf0448890 from core warning: no source compiled with -g [using memory image in core] warning: Unable to access address 0xf03488e4 from core Segmentation fault in is_data_there at 0xd020e5fc 0xd020e5fc (is_data_there+0x1c) 9421ffc0 stwu r1,-64(r1) (/usr/ccs/bin/dbx) where is_data_there() at 0xd020e5fc __which_res_state() at 0xd020eacc ho_byname2(??, ??, ??) at 0xd022bbd8 ho_byname2(??, ??, ??) at 0xd0241ef0 gethostbyname2(??, ??) at 0xd020dc34 dowprnt.mbcounter() at 0xd173b4e8 g_timer_new() at 0xd1af14b0 g_timer_new() at 0xd1af14f8 g_timer_new() at 0xd1af362c dowprnt.mbcounter() at 0xd1625b30 dowprnt.mbcounter() at 0xd1729e54 pthread._pthread_body(??) at 0xd00080c8 ------------------------------------------------------ new profile: ------------------------------------------------------ $ /opt/mozilla/mozilla -profileManager /opt/mozilla/run-mozilla.sh[36]: 28962 Segmentation fault(coredump) Oh no! /opt/mozilla/mozilla-bin just dumped a core file. Do you want to debug this ? You need a lot of memory for this, so watch out ? [y/n] y ...skip... warning: Unable to access address 0xf0448890 from core warning: Unable to access address 0xf0448890 from core warning: no source compiled with -g [using memory image in core] warning: Unable to access address 0xf03488e4 from core Segmentation fault in is_data_there at 0xd020e5fc 0xd020e5fc (is_data_there+0x1c) 9421ffc0 stwu r1,-64(r1) (/usr/ccs/bin/dbx) where is_data_there() at 0xd020e5fc __which_res_state() at 0xd020eacc ho_byname2(??, ??, ??) at 0xd022bbd8 ho_byname2(??, ??, ??) at 0xd0241ef0 gethostbyname2(??, ??) at 0xd020dc34 dowprnt.mbcounter() at 0xd173b4e8 g_timer_new() at 0xd1af14b0 g_timer_new() at 0xd1af14f8 g_timer_new() at 0xd1af362c dowprnt.mbcounter() at 0xd1625b30 dowprnt.mbcounter() at 0xd1729e54 pthread._pthread_body(??) at 0xd00080c8 ------------------------------------------------------ rm -rf ~/.mozilla ~/.netscape ------------------------------------------------------ $ /opt/mozilla/mozilla /opt/mozilla/run-mozilla.sh[36]: 29078 Segmentation fault(coredump) Oh no! /opt/mozilla/mozilla-bin just dumped a core file. Do you want to debug this ? You need a lot of memory for this, so watch out ? [y/n] y warning: Unable to access address 0xf0448890 from core warning: Unable to access address 0xf0448890 from core warning: no source compiled with -g [using memory image in core] warning: Unable to access address 0xf03488e4 from core Segmentation fault in is_data_there at 0xd020e5fc 0xd020e5fc (is_data_there+0x1c) 9421ffc0 stwu r1,-64(r1) (/usr/ccs/bin/dbx) where is_data_there() at 0xd020e5fc __which_res_state() at 0xd020eacc ho_byname2(??, ??, ??) at 0xd022bbd8 ho_byname2(??, ??, ??) at 0xd0241ef0 gethostbyname2(??, ??) at 0xd020dc34 dowprnt.mbcounter() at 0xd173b4e8 g_timer_new() at 0xd1af14b0 g_timer_new() at 0xd1af14f8 g_timer_new() at 0xd1af362c dowprnt.mbcounter() at 0xd1625b30 dowprnt.mbcounter() at 0xd1729e54 pthread._pthread_body(??) at 0xd00080c8 ------------------------------------------------------
Trying to reproduce problem. Unable to recreate on AIX 4.3.3 ML09 - testing on ML10 with latest fixes.
Mozilla 1.2.1 comes up in AIX 4.3.3 ML10. I will incrememtally upgrade bos.rte.libc until I can recreate this problem. A similar problem has been reported in Netscape 7 for AIX with libc 4.3.3.88.
Status: UNCONFIRMED → NEW
Ever confirmed: true
Taking ownership of bug.
Assignee: asa → pkw
I have confirmed that this is occurring after the upgrade of bos.rte.libc from 4.3.3.87 to 4.3.3.88. Here is a stack trace from a debug build of Mozilla 1.2.1 (provides more information): aix_thread_info.__which_res_state() at 0xd020e7d4 gen_nw.irs_gen_nw(??) at 0xd0241c10 gethostent.gethostbyname2(??, ??) at 0xd020d954 prnetdb.PR_GetIPNodeByName(0x20d571d8, 0x180018, 0x30, 0x20d5744c, 0x4a4, 0x20d57438), line 765 in "prnetdb.c" nsDnsService.DoSyncLookupInternal()(0x20d57418), line 857 in "nsDnsService.cpp" nsDnsService.DoSyncLookup()(0x20d57418), line 871 in "nsDnsService.cpp" unnamed block nsDnsService.Run().$b40240, line 1403 in "nsDnsService.cpp" nsDnsService.Run()(0x200dc0a8), line 1403 in "nsDnsService.cpp" nsThread.Main(void*)(0x200dc308), line 120 in "nsThread.cpp" ptthread._pt_root(0x200dc518), line 214 in "ptthread.c" pthread._pthread_body(??) at 0xd00080c8
Status: NEW → ASSIGNED
bos.rte.libc 4.3.3.88 includes a fix for a security bug which greatly increases the amount of stack space used when calling functions like gethostbyname. Mozilla creates threads with stack size 64k on AIX - this combined with the new stack requirements of the latest libc cause this crash to occur. I have suggested an alternate fix for the bos.rte.libc fileset, but for now I will post a temporary workaround which increases the stack size on AIX in NSPR. It should only be necessary to drop in a new libnspr4.so for Mozilla to work with the new libc.
Attachment #110785 - Flags: review?(wtc)
Comment on attachment 110785 [details] [diff] [review] Increases AIX stack from 64k to 128k r=Roland.Mainz@informatik.med.uni-giessen.de
Attachment #110785 - Flags: superreview?(wtc)
Attachment #110785 - Flags: review?(wtc)
Attachment #110785 - Flags: review?(Roland.Mainz)
Attachment #110785 - Flags: review?(Roland.Mainz) → review+
Comment on attachment 110785 [details] [diff] [review] Increases AIX stack from 64k to 128k This patch is fine.
Attachment #110785 - Flags: superreview?(wtc) → superreview+
wtc: Can you check the patch "in" into "trunk" (and 1.2.1-branch if possible), please ?
I've checked in the patch on the NSPR trunk and NSPRPUB_PRE_4_2_CLIENT_BRANCH (used by Mozilla trunk). I don't know how to get a fix checked into Mozilla 1.2.1. Someone else needs to get the approval... With the approval and the cvs branch it should go into, I'll be happy to check it in.
This is fixed on the trunk. I don't plan on backporting this to the 1.2.1 build.
Status: ASSIGNED → RESOLVED
Closed: 22 years ago
Resolution: --- → FIXED
Verified on AIX machine with bos.rte.libc 4.3.3.88 applied.
Status: RESOLVED → VERIFIED
Philip K. Warren wrote: > This is fixed on the trunk. I don't plan on backporting this to the 1.2.1 > build. Is it possible to _refresh_ the 1.2.1 AIX release build with the updated NSPR library ?
*** Bug 189086 has been marked as a duplicate of this bug. ***
This problem can be resolved by applying the following APARS when they become available: AIX 4.3: IY39444 AIX 5.1: IY39429 AIX 5.2: IY39589 These address the bad fix to bos.rte.libc which caused these crashes to occur. In addition, I have posted an updated Netscape 7 (VRMF: 7.0.0.1) to the AIX Web Browsers page which resolves this issue in Netscape 7. It can be downloaded from the following location: http://www14.software.ibm.com/webapp/download/search.jsp?go=y&rs=netscape7
Product: Browser → Seamonkey
You need to log in before you can comment on or make changes to this bug.

Attachment

General

Creator:
Created:
Updated:
Size: