[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]
Re: lynx-dev coredump
From: |
dickey |
Subject: |
Re: lynx-dev coredump |
Date: |
Sat, 31 Oct 1998 09:06:28 -0500 (EST) |
>
> >> but now come again). What should I do to prepare a debug info for you?
> >> I got exit with signal 11 on Linux.
>
> > well - if I were seeing coredumps in something, I'd build a debug-version
> > (compile with -g) so that I could see the symbols in the core file better.
> > That's a start. (If it is my own machine, I'd link with ElectricFence,
> > or at work, with Purify, so that I can get a precise location of the
> > problem - but both have limitations, and you can't generally run those
> > on your ISP). Then I'd simply run with the debug version until the problem
> > came back (if you're running the same optimization level, etc., it probably
> > will).
>
> Ok, I build a debug version of 2.8.1rel.1,
> link it with ElectricFence, and run under gdb.
> Apperently, the problem around freeing the "title" in LYpop(), *sometimes*.
> Any ideas?
perhaps we're freeing the "title" pointer twice.
> (gdb) bt
> #0 0x400709e9 in __kill ()
> #1 0x80c974b in EF_Abort ()
> #2 0x80c9269 in free ()
> #3 0x8077a81 in LYpop (doc=0x810d6e4) at ./LYHistory.c:312
> #4 0x805e4df in mainloop () at ./LYMainLoop.c:381
> #5 0x805cf98 in main (argc=1, argv=0xbffffc58) at ./LYMain.c:1686
> #6 0x804a65b in ___crt_dummy__ ()
> (gdb) up
> #1 0x80c974b in EF_Abort ()
> (gdb) up
> #2 0x80c9269 in free ()
> (gdb) up
> #3 0x8077a81 in LYpop (doc=0x810d6e4) at ./LYHistory.c:312
> 312 FREE(doc->title);
> (gdb) up
> #4 0x805e4df in mainloop () at ./LYMainLoop.c:381
> 381 LYpop(&newdoc);
> (gdb) up
> #5 0x805cf98 in main (argc=1, argv=0xbffffc58) at ./LYMain.c:1686
> 1686 status = mainloop();
> (gdb) up
> #6 0x804a65b in ___crt_dummy__ ()
> (gdb)
>
>
>
> > Then - in addition to a walkback (giving line numbers and parameter
> > values),
> > look at the values of local symbols to try to pinpoint _why_ the program
> > broke. (If it's a null-pointer, or a trashed value, for instance).
>
> > A walkback from an executable that's in assembler-format is sometimes
> > useful, but I've not learned how to match the arguments shown by adb
> > properly against the C code.
>
> >> Apperently, I got a last coredump when was printing your message to
> >> e-mail.
> >> I receive the message OK but its subject happens to be from Print menu!!!
> >> This may be a starting point for investigations...
>
>
--
Thomas E. Dickey
address@hidden
http://www.clark.net/pub/dickey