A patch for libio/tst-widetext.c
Jakub Jelinek
jakub@redhat.com
Tue Aug 1 13:27:00 GMT 2000
On Tue, Aug 01, 2000 at 01:22:30PM -0700, H . J . Lu wrote:
> On Tue, Aug 01, 2000 at 10:13:16PM +0200, Jakub Jelinek wrote:
> > On Tue, Aug 01, 2000 at 01:04:05PM -0700, Ulrich Drepper wrote:
> > > Mark Kettenis <kettenis@wins.uva.nl> writes:
> > >
> > > > This is with Linux 2.2.14, glibc 2.1.3 (not the official release but
> > > > something from CVS I installed on June 12), bash 2.04.0(2) (I believed
> > > > I patched that one to get rid if the famous `invalid character 45'
> > > > problem) and a make I compiled myself straight from the tar file.
> > > >
> > > > Looks like RedHet screwed this up.
> > >
> > > I've looked through the changes from 3.79 to 3.79.1 and the problem is
> > > that make tries to use the clock_*() functions which are new in 2.2
> > > (therefore you don't see the problem with glibc 2.1.3). The clock_*()
> > > functions are in librt which brings in libpthread which restricts the
> > > stack size to 2MB.
> >
> > Yes, I just got through the same investigation.
> > I think I'll just disable clock_* checking in the configure script for now.
>
> Hi, Jakub,
>
> Just a thought. Is it related to the ntp problem which you mentioned
> it has something to do with clock_*() functions?
Not to my knowledge. make seemed to work well with clock_* functions in,
unlike ntp, so it is highly probably it is just ntp using bad code when
those functions are available. Unfortunately, it is quite hard to debug,
because it only happens when it is not under the debugger, is dettached from
tty etc. For now, we use the same hack which I did now for make - pretend no
clock_* in configure.
Jakub
More information about the Libc-hacker
mailing list