[Date Prev][Date Next] [Thread Prev][Thread Next] [Date Index] [Thread Index]

Re: Native Netscape runs on Debian-sparc (And glibc patches)



On Sun, Dec 20, 1998 at 11:27:26AM -0500, Steve Dunham wrote:
> Christian Meder <meder@isr.uni-stuttgart.de> writes:
> 
> > Sorry for the late answer. I had a pretty busy week.
> 
> Sorry for jumping the gun on releasing glibc...
> 
> > > Christian, are you planning on upgrading the Debian sparc distribution
> > > to glibc-2.1_2.0.105 anytime soon?  If so I have an essential patch
> > > for you.
> 
> > I planned to update glibc but my experiences with glibc2.1 taught me
> > to be careful about it. Anyway you are not stepping on my toes by
> > doing an NMU. Remember just be careful and double check it ;-)
> 
> I'm rebuilding with patch-sigaction.  I noticed that "patch-ldso" is
> also not enabled.  Do you have any way to test if we still need it
> (it's a fairly ugly hack - it ifdef's out some code because "__getcwd
> Segfaults on 2.0.35").
> 
> I restarted the build with the patch-ldso included.  I saw nothing in
> the glibc changelog about getcwd, so I assumed that it is still
> needed. Do you have any more details on this problem?

It's a pretty weird problem. getcwd is a syscall in 2.1.x kernels so
everything said only applies to 2.0.x kernels. The facts:

* the getcwd function works fine on 2.0.x kernels. I tried some simple
C programs. Besides it's tested during the make test phase of the
glibc2.1 build
 
* this specific getcwd invocation in the dynamic linker segfaults
reproducibly (at least for me). It is triggered by executing something
like
LD_LIBRARY_PATH=<some _relative_ path> <some program>
with a glibc2.1 without the patch. If you use a relative path in one
of the dynamic linker environment variables the linker tries to lookup
the full path with getcwd and segfaults. The ifdef causes the linker
to skip the getcwd and just use the relative path (which isn't too
annoying I hope)

I guess the problem is somewhere hidden in egcs but I don't
know. Unfortunately this ugly hack was pretty important because a lot
of packages set the LD_... variables during the build and the
segfaults broke the build processes.

Greetings,


				Christian

PS.: Sorry to the list for this boring technical stuff ;-) 
-- 
Christian Meder, email: meder@isr.uni-stuttgart.de

What's the railroad to me ?
I never go to see
Where it ends.
It fills a few hollows,
And makes banks for the swallows, 
It sets the sand a-blowing,
And the blackberries a-growing.
                      (Henry David Thoreau)


Reply to: