On Sat, Jul 24, 2010 at 5:29 PM, Dawe <dawed...@gmx.de> wrote:
> On Sat, 24 Jul 2010 12:30:56 +0200
> Dawe <dawed...@gmx.de> wrote:
>
>> On Fri, 23 Jul 2010 20:31:54 -0700
>> patrick keshishian <pkesh...@gmail.com> wrote:
>>
>> > On Fri, Jul 23, 2010 at 12:26 PM, Dawe <dawed...@gmx.de> wrote:
>> > > On Thu, 22 Jul 2010 22:16:32 -0700
>> > > patrick keshishian <pkesh...@gmail.com> wrote:
>> > >
>> > >> Seems at least the last two macppc snapshots I've tried had a broken
>> > >> X. I saw mention of this from another user on po...@[1].
>> > >>
>> > >> Just curious if this is being worked on, is there anything I can do to
help?
>> > >>
>> > >>
>> > >> [1] http://marc.info/?l=openbsd-ports&m=127983358723897&w=2
>> > >>
>> > >> [2] dmesg copy-and-pasted from notepad.exe (sorry for any mangling)
>> > >>
>> > >
>> > > My own build of X works again.
>> > > Maybe something just in snapshots?
>> >
>> > Hmm... after I rebuild X again it seems to work now (still some issues
>> > though). I'm curious, do you know which snapshot you were using?
>> >
>> > So just for the record:
>> > - X from snapshots (July 19th, and July 22nd) did not work.
>> >   With the July 19th snapshot I did a fresh install. With the July 22nd
>> >   snapshot I took the upgrade path.
>> > - Building X from source (empty /usr/xobj, etc) still did not work.
>> > - Built xserver with debug and following instructions in xenocara/README
>> >   left me with a strange core with useless backtrace.
>> > - Rebuilding X (empty /usr/xobj) with DEBUG=-g starts X but there
>> >   are some issues with some ports that use X (e.g., firefox 3.6.7
>> >   crashes similar useless core file as X earlier).
>> >
>> > ATM I'm starting to build the entire system from source (kernel, base,
>> > X over again) and then try and (re)build the ports ...
>> >
>> > --patrick
>> >
>>
>> The snapshot was from July 22nd. That was the first time I noticed that
>> X was broken. I can't say when it worked the last time because I don't
>> use X on this machine regularly.
>>
>> I build X with env CFLAGS=-g make build, it was the first X build on
>> this machine, so xobj was clean.
>>
>> I'm seeing similar crashes like you with firefox. Other X using ports
>> don't seem to have such issues. gvim, midori, gkrellm, qemu all work so
>> far.
>>
>> I'll build a firefox package with debug symbols to see how this works
>> out.
>>
>
> Okay, I rebuild firefox35 with -g.
> Seems to be a lib loading problem.
>
> (gdb) bt full
> #0  0x265052f8 in dlsym () from /usr/libexec/ld.so
> No symbol table info available.
> #1  0x2653af94 in ?? () from /usr/libexec/ld.so
> No symbol table info available.
> #2  0x2653af94 in ?? () from /usr/libexec/ld.so
> No symbol table info available.
> Previous frame identical to this frame (corrupt stack?)

That is very similar stack trace to what I got with both X[1] and
firefox 3.6.7 after getting X working (by compiling it with
CFLAGS=-g).

> Running firefox35-bin in gdb:
> (gdb) run
> Starting program: /usr/local/firefox35/firefox35-bin
> /usr/local/firefox35/firefox35-bin: can't load library 'libxul.so.23.0'

I assume ldd is able to find the library?

--patrick
[1] http://marc.info/?l=openbsd-misc&m=127990972425965&w=2

Reply via email to