Re: depcomp testing

[Alexandre Duret-Lutz]

Sorry for the delay,

>>> "RW" == Ralf Wildenhues <address@hidden> writes:

[...]

 RW> Below, there's also a hack for depmodes hp/gcc to not output `foo.c :'
 RW> lines.  It's a hack because `grep -v "^$source$"' is bad when $source
 RW> contains regex-active characters, but was the simplest I could come up
 RW> with.  Better suggestions welcome.

It's not clear to me why we should bother, especially since the
correction is fragile extra work.  The only situation I could
think of where the extra line mights change something is if you
delete the said file without removing its name from the Makefile.am, but
then even if make won't complain about the missing file the compiler
will so this (IMHO marginal) error won't be missed.

The extra test look useful even without this point, so please go
ahead and check them.

 RW> * tests/depcomp6.test, tests/depcomp7.test: New tests,
 RW> for general `depcomp' functionality, with and without
 RW> `subdir-objects', with and without `libtool'.
 RW> * tests/Makefile.am: Adjusted.

"Adjust."  (The ChangeLog uses imperative.)

[...]

 RW> +  sleep 1

Just use $sleep (see tests/defs.in).  

make maintainer-check should probably warn about this, although I haven't run
it for a while

 RW> +  if $MAKE; then false; fi

Let's write "$MAKE && exit 1" because that's what of the other test do, and
I'm wary that some shell will return the $? of $MAKE if it fails, in turn
causing "set -e" to abort the script.  Don't ask me which shell, but that's 
why we have "else :; fi" all over the Makefiles.
-- 
Alexandre Duret-Lutz

Shared books are happy books.     http://www.bookcrossing.com/friend/gadl