This is the mail archive of the
libc-alpha@sources.redhat.com
mailing list for the glibc project.
Re: Spurious rebuild of Versions.v.i
- From: Daniel Jacobowitz <drow at mvista dot com>
- To: libc-alpha at sources dot redhat dot com
- Date: Sun, 2 Nov 2003 00:49:55 -0500
- Subject: Re: Spurious rebuild of Versions.v.i
- References: <20031030165447.GA27589@nevyn.them.org> <200310302153.h9ULrYNR018089@magilla.sf.frob.com> <20031102043847.GA13326@nevyn.them.org>
On Sat, Nov 01, 2003 at 11:38:47PM -0500, Daniel Jacobowitz wrote:
> On Thu, Oct 30, 2003 at 01:53:34PM -0800, Roland McGrath wrote:
> > > I'm seeing a really strange build failure. It's too intermittent for me to
> > > reproduce with debugging.
> >
> > If you can handle the spew, you could try passing --debug=b to make all the
> > time so we can examine the messages the next time it bites you.
> >
> > > Something causes Versions.v.i to be rebuild during processing of a
> > > subdir; unfortunately, it happens during resolv.
> >
> > We should figure out and fix the spurious rebuild problem. But you could
> > have changed shlib-versions or whatever and then rebuilt in resolv and that
> > should work right. Setting CPPFLAGS globally in a subdir is highly
> > questionable. For resolv, I think that setting is really only needed for
> > gethnamadr.c and so it should be CPPFLAGS-gethnamadr.c instead.
>
> It looks to me as if it's some sort of nanosecond-timestamps problem.
> I'll run a couple of --debug=b builds to see if I can catch it in
> action, but a frustrated evening chasing it down didn't turn up a
> cause. It's causing all sorts of bizarre errors; for instance it
> caused pthread_create to be temporarily referenced as an external
> symbol from librt.so, causing link failures in make check; five minutes
> later librt.so was rebuilt and the error vanished. It also caused
> libc.so to be relinked during make check, causing temporary failures
> for tests running in parallel that tried to link to it.
Also included are some non-parallel-make-safe constructs. In general,
anything which can cause recursion to a parent directory is unsafe; I
just had two subdirectories attempt to recreate Versions.v.i at the
same time. Looks like rtld-csu and rtld-gmon.
I know this isn't terribly helpful but I don't have time to track down
the problem with any more depth right now.
--
Daniel Jacobowitz
MontaVista Software Debian GNU/Linux Developer