Red Hat Bugzilla – Bug 98768
Core dump and segfaults
Last modified: 2007-04-18 12:55:33 EDT
From Bugzilla Helper:
User-Agent: Mozilla/5.0 (X11; U; Linux i686; en-US; rv:1.3.1) Gecko/20030527
Description of problem:
Note: I am here only mimicking with mandrake but the bug exists in redhat 9 .It
has been verified many a times and i found that bug
exists in Redhat 9.So there must be a common solution.
I recently installed viavoice tts .I thought that the bug below was
caused due to errors in viavoice but later when i did some searching i found
that it is due to glibc 2.3 which segfaults during the execution.I arrived
at this conclusion of glibc because i tried debian glibc 2.3 which works fine
and the redhat glibc 2.3 (from 9.0 installation) is causing the problem.
So to make sure that bug is due to redhat glibc 2.3 i installed the viavoice
in redhat 7.3.It worked well.Later upon searching the net i couldn't find
much info about how to correct the problem until i read a post saying that
it was a bug due to glibc redhat patch.So what should i do ? Update glibc
or can it be worked around without updating glibc .
The "###DEBUG Hi1" was added by me for finding out which function is causing
to segfault.Unfortunately it is the eciNew() function.
Also please tell me what other info should i attach along with bugs.What are
the relevant things which will help you to solve the problem more easily.
More info at this url .I posted it once but it was giving me URI too long
error.So here is the link.
Version-Release number of selected component (if applicable):
Steps to Reproduce:
1.Install viavoice tts rpm
3. See the segfault
rpm -ivh viavoice
Actual Results: It crashed and core was produced
Expected Results: ./cmdlinespeak should run. Instead it segfaults dumps core
Please see the description
Well, first of all, we issued glibc errata for RHL9, so you should be testing
with glibc-2.3.2-27.9, not -11.9.
From the backtrace it looks like the application uses the braindamaged
To debug this, I need to know:
a) what arguments has the compile() supposedly in /usr/lib/enu50.so
been called with (if more times, then the one which uses the same expbuf
as the segfaulting advance is interesting)
b) what arguments have been passed to regcomp from the compiled function
c) what arguments have been passed to advance when it segfaulted
a) compile takes 1 string, 1 expbuf ptr, 1 endbuf ptr and 1 intereger,
so if you insert breakpoint on the first insn in it you can just
x/5w $esp and x/1s *(char **)($esp + 4) in gdb
b) regcomp takes 1 expbuf ptr, 1 string with regular expression, 1 integer
again, x/4w $esp, x/1s *(char **)($esp + 8) in gdb
c) advance takes 1 string and 1 expbuf ptr, so
x/3w $esp, x/1s *(char **)$(esp + 4)
Ping. Did you try the current errata glibc? And more importantly, did you look
at IBM for updates ti this program? Given the highly suspicious interfaces they
are using I put my money on an application bug.
No response in a month. I can see no bug in glibc. Closing now.