Re: 02-distclean-test.diff

[Alexandre Duret-Lutz]

>>> "Derek" == Derek Robert Price <address@hidden> writes:

 Derek> Okay, this patch includes the fix.  It seems to work correctly, except
 Derek> multiple invocations of AC_CONFIG_LINKS with the same source file seem
 Derek> to cause multiple warnings when the file is missing:

 Derek> configure.in:10: required file `./sdir/src' not found
 Derek> configure.in:10: required file `sdir/src' not found

 Derek> Should I worry about this?  Would you like me to write
 Derek> another test for the require functionality?

I don't think you should worry about it.  That's probably due to
the strange logic in the require_file* functions.  Rewriting
these have been on the TODO list for a very long time.

Another facet of the require functionality you could check, if
you want, is that these files actually get distributed.  (I like
to do this from a make rule, so that I don't have to know
$(distdir) -- for example see include2.test.)

 Derek> One last question, why is only the last $where preserved
 Derek> for AC_CONFIG_FILES and the like?  

I can't think of any reason, unless all that code was copied
from AC_CONFIG_HEADER which never really supported more than one
header (due to the rebuild rules that call autoheader, IIRC).

 Derek> I duplicated this behavior for AC_CONFIG_LINKS, but it
 Derek> doesn't seem like it would be very hard to preserve the
 Derek> exact $where for each macro invocation.

Indeed.

 Derek> Index: automake.in

Looks great.

[...]

 Derek> +. ./defs || exit 1
 Derek> +
 Derek> +echo 'SUBDIRS = sdir' > Makefile.am
 Derek> +: > src
 Derek> +mkdir sdir
 Derek> +: > sdir/Makefile.am
 Derek> +: > sdir/src
 Derek> +mkdir sdir-no-make
 Derek> +
 Derek> +cat >>configure.in << 'EOF'
 Derek> +AM_INIT_AUTOMAKE

You shouldn't need this one: it's already in configure.in.

(Aside: I'm surprised one can call AM_INIT_AUTOMAKE twice.
Maybe we should have a sanity check about this.)

 Derek> +AC_CONFIG_FILES(sdir/Makefile)
 Derek> +AC_CONFIG_LINKS(dest:src)
 Derek> +AC_CONFIG_LINKS(dest2:src)
 Derek> +AC_CONFIG_LINKS(sdir/dest:src)
 Derek> +AC_CONFIG_LINKS(dest3:sdir/src)
 Derek> +AC_CONFIG_LINKS(sdir/dest2:sdir/src sdir-no-make/dest:src)
 Derek> +AC_OUTPUT
 Derek> +EOF
 Derek> +
 Derek> +set -e

Better put `set -e' directly after ./defs || exit 1.  Even the
echo/mkdir/:/cat can fail.

[...]

 Derek> +set +e

I think this is not needed either.  `false && exit 1' is
perfectly safe with `set -e'.  

 Derek> +test -r dest && exit 1
 Derek> +test -r dest2 && exit 1
 Derek> +test -r sdir/dest && exit 1
 Derek> +test -r dest3 && exit 1
 Derek> +test -r sdir/dest2 && exit 1
 Derek> +test -r sdir-no-make/dest && exit 1

I'm surprised this succeeds.  When sdir-no-make/dest does not
exist (which is expected), this last line should return with $?
= 1.  At least that's how it works with all the shell I've tried
(independently of the set-e/set+e setting).  Because it's the
last command, the whole script will exit with this same $?,
hence reporting a failure.

One generally adds a `exit 0', or simply `:', at the end of a
test when the last check is negated.
-- 
Alexandre Duret-Lutz