Re: PATH_MAX
On Mon, Dec 17, Christopher Yeoh wrote:
> Thorsten Kukuk writes:
> >
> > So a lot of tests fails.
> > If you know uses a fixed kernel and recompiled glibc with
> > PATH_MAX = 4096, also some tests fails, because sysconfig
> > does not return the expected value.
> >
> > Any suggestions what to do? Should LSB require PATH_MAX 4096
> > and compile the test suite with this?
>
> I believe we were planning on waiving these test failures due to this
> PATH_MAX problem as fixes for the kernel and glibc and still flowing
> through the systems. But it should be 4096 and it would preferable to
> have tests pass where this has been corrected.
>
> I think I can just compile the next release of the binary test suites
> with PATH_MAX set to 4096 (does glibc need any other patches for this
> to work properly?).
I haven't audited the glibc code complete yet, but it looks like it
is enough to recompile glibc with PATH_MAX = 4096. You need your own
copy of kernel includes in /usr/include (like SuSE Linux has) and
to modify /usr/include/linux/limits.h. In userland we expect in all
cases I checked yet that PATH_MAX contains the 0.
> This will lead to some other tests failing on systems where PATH_MAX
> has not been fixed due to sysconfig inconsistencies but we can waive
> these for a limited time.
If you recompile glibc and then the test suite with the new header, there
are no new tests which fails and the PATH_MAX tests passes all. At least
for me. You don't need to patch the kernel, because the kernel
calculates PATH_MAX + trailing 0, which is 4096.
Thorsten
--
Thorsten Kukuk http://www.suse.de/~kukuk/ kukuk@suse.de
SuSE GmbH Deutschherrenstr. 15-19 D-90429 Nuernberg
--------------------------------------------------------------------
Key fingerprint = A368 676B 5E1B 3E46 CFCE 2D97 F8FD 4E23 56C6 FB4B
Reply to: