Re: Second (non-fPIC) pass messages being suppressed on failure

[Ed Catmur]

On Thu, 2007-11-22 at 08:19 +0100, Ralf Wildenhues wrote:
> >  However, if the second position-dependent run fails
> > (compiler bug, usually, but could also be running out of disk space) it
> > will fail without any error message being printed.
> 
> Yes.  Have you actually hit a compiler bug that way before?  I think
> it's much more likely that the set of warnings is different rather
> than a failure, but others may have a different experience.

Yes; the issue that prompted this was an incident[1] where the -fPIC
pass compiled without any warnings at all, while the non-fPIC pass was
generating a spurious warning[2] - that wouldn't have been a problem,
but the maintainers had decided that the package should always be
compiled with -Werror.

It's a tiny corner case, I know - but if I've been bitten then certainly
other people will have seen this and not had the patience to debug it.

> > Would it make sense to capture output from the second compilation
> > instead of discarding it, then print it if the second compilation fails?
> 
> Suggested patch against HEAD.  It costs one more fork of `rm' per
> compile, and makes the typical echoed command line longer, but saves
> the user from having to type something like
>   make || env CFLAGS='... -no-suppress' make -e
> 
> WDYT?  (Adding Ed to THANKS not shown.)

That looks great, thanks!

Ed

1. http://bugs.gentoo.org/show_bug.cgi?id=196103
2. http://gcc.gnu.org/bugzilla/show_bug.cgi?id=29478