FC5 breaks Accelrys InsightII - details

[Jakub Jelinek]

On Fri, Apr 07, 2006 at 11:45:02AM -0400, Jack Howarth wrote:
> Jakub, would it be a reasonable from this observation to draw the
> conclusion that the bulk of InsightII was built on a properly
> configured build machine but that specific libraries must be
> stale old copies from much older build? Otherwise it doesn't make
> much sense as to why libcblib.so and libchemnote.so shared the
> same libraries in their linkage yet one has both the incorrect and
> correct errno symbol in it.

If you googled around on this topic, you'd find many answers.
The bug in some input files linked into the buggy library is that
it declares errno itself, like with
extern int errno;
rather than doing what standards require, particularly:
#include <errno.h>
If you do this with ~ RHL9+ glibc, the resulting library won't link,
you'll get linker error, previously this bug wasn't diagnosed
(though ever since introduction of threads into glibc it had no chance
of working properly in threaded environment).

	Jakub