Antwort: Re: [uClibc] PowerPC 0.9.22 ldso and dynamic libstdc++ still no-go

[Richard_Theil at dadebehring.com]

On Thu Nov 13, 2003 at 11:21:29AM +0100, Richard_Theil at dadebehring.com 
wrote:

Just tried Joakims last patch:

-                                insns[0] = OPCODE_B(delta - 4);
+                                insns[0] = OPCODE_B(delta);

On my 0.9.22-plus-4-CVS-files tree. Same old crash.

>> ...I can't access CVS from behind the corporate proxy,
>> so I got all the recent ldso files involved over the web

> Why not just grab the latest nightly snapshot?  I would
> imagine that would be quite a lot easier....

This shop appreciates work on minimal (and more important
reproducible) changesets. But I guess as long as the issues
persist, I might as well grab the snapshot. Yes.

>> The symptoms are exactly as before: SIGSEGV with 8 in r31
>> at the same location.
>Hmm.  uClibc shared libs do not work at all?  Or libstdc++
>doesn't work, but uClibc shared libs work?

I described that in my original posting (from 2003-09-17,
that even contains the source for my little test, which I
built with "powerpc-uclibc-g++ -o hellopp hellopp.cpp" ):

- C works just fine, shared or not
- C++ works just fine when I build with --static
- C++ crashes during what looks to be libstdc++
  initialization when it is built normally
  (= dynamically linked).

I have not experimented with complicated static C data,
so I can't really tell whether it is mis-loading (I'd
assume the loader gets a large zeroed area right), or
something else.

> The more specific you can be, the more likely it is
> we can be helpful.

I'm deeper in this now than I ever wanted to be, becoming
good friends with objdump and readelf :( so once I figure
out where the loader places the segments, I can let you
know at what segment/offset I encounter bogus data.
__SUPPORT_LD_DEBUG__ might help too... Might take a while
though, because there's more important work queueing up
here as long as the static link works for now.

Rich

-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://lists.busybox.net/pipermail/uclibc/attachments/20031113/181ffbbd/attachment.htm