Nelson,

I am absolutely not saying that the problem is in NSS. I posted the first 
problem to opensc's maillist when I did it here for the first time. I was 
actually suspecting that problems could exist in opensc's pkcs11 module 
because before the latest version (0.7.0), I had, for example, problems 
when having Mozillas master password set to "only first time needed". 
Desafortunately, I haven't got any answer from the opensc guys.

You know my first post was about problems getting right cert when 
inserting/removing smart cards (issue 1). The second one was calling 
PK11_ListCerts, signing and calling again (issue 2). Ok, I will concentrate 
in this last issue.

I downloaded and built Mozilla 1.0 from source. Actually I did not need to 
debug anything as issue 2 is not an issue in 1.0. I get no problems. It is 
working well.

Then I got nightly source (Mozilla 1.3a - Mozilla/5.0 (X11; U;
>> Linux i686; en-US; rv:1.3a) Gecko/20021119) and built as well. The goal 
was to debug and provide you with stack trace. I have tried and tried but I 
can't get it work. My first idea was to set a breakpoint like " b _dl_open" 
to see when either nss3 or my plug-in was loaded. Knowing that nss is 
loaded (info shar), I could then do " b PK11_ListCerts" and after that step 
to see where the crash happens.

But I get problems with a simpler thing like:

$ gdb ./mozilla -g
(gdb) b main
Breakpoint 1 at 0x805f734: file nsAppRunner.cpp, line 1796.
(gdb) r
Starting program: /home/rago/software/mozilla/dist/bin/./mozilla-bin
[New Thread 1024 (LWP 26816)]
[Switching to Thread 1024 (LWP 26816)]

Breakpoint 1, main (argc=1, argv=0xbffff2b4) at nsAppRunner.cpp:1796
1796      InstallUnixSignalHandlers(argv[0]);
Current language:  auto; currently c++
(gdb) c

as I get:
---------

Continuing.
Assertion failure: 0.5 <= maxAlpha && maxAlpha < 1 && 0 <= minAlpha, at 
pldhash.c:227

Program received signal SIGABRT, Aborted.
0x406b2861 in kill () from /lib/libc.so.6


I don't know what's going on. I tried to mask the assertion via gdb handle 
SIGABRT (not passing to program and not stopping), but finally after one 
more assertion warning, I got a segmentation fault.

Hmm, so far, the only thing that it is a bit clear for me is that Mozilla 
1.0 is working better with opensc-0.7.0 than Mozilla 1.3.a (also than 11b, 
the only ones I tried).

I am very interested in finding where the problem is. I would like to 
provide you with a good stack trace, but I need some help/hints about how 
to avoid the problem above with gdb.

Best regards,
/R




Nelson B. Bolyard wrote:

> raffe wrote:
> 
>> I have tried with latest nightly [Mozilla 1.3a - Mozilla/5.0 (X11; U;
>> Linux i686; en-US; rv:1.3a) Gecko/20021119].
>> 
>> Sorry, but even worse! I call PK11_ListCerts, I sign with a cert in smart
>> card, I call PK11_ListCerts again and I get a crash ! This is worse than
>> what made me write to you guys (removing/re-inserting cards).
>> 
>> I can call PK11_ListCerts so many times I want without problems if I do
>> not sign anything. Both soft & hard certs are correctly got. But, as soon
>> as I sign with one of the certs (no matter if soft or hard), then, next
>> time I call PK11_ListCerts, I get a big crash.
>> 
>> Another annoying thing is that the first time Mozilla crashed, my soft
>> certs dissappeard from the database.
> 
> Right.  But we don't yet know whether the crashes are occurring in NSS or
> in the OpenSC module.  Until we have at least a stack trace of the crash,
> I wouldn't assume the problem is in NSS.
> 
> If the problem is worse in newer builds, that only means the code is
> tickling the bug sooner, but still doesn't tell us whether it's in NSS or
> the PKCS11 module.
> 
> --
> Nelson Bolyard
> Disclaimer:                  I speak for myself, not for Netscape


Reply via email to