Re: 10-check-am.patch

[Akim Demaille]

| >>>>> "Akim" == Akim Demaille <address@hidden> writes:
| 
| Akim>         * automake.in (handle_phony): Sort.
| Akim>         * automake.in (file_contents_with_transform): Catch `.PHONY:'.
| Akim>         * check.am: New file.
| 
| check.am wasn't in the patch.  However, I trust it is just an
| extraction from handle_tests.

Arg...  I must have left it at home too :(  But indeed, it's _almost_
just like what was in handle_tests but the fact that I included

.PHONY: check-TESTS

This is the most debatable part of my patches, and the one which
certainly needs rewriting.  In fact I was forcing myself to implement
it to see if I would `feel' something better.

And here is what I would suggest: specify a list of dependencies to
factor, and have file_contents_with_transform always factor these.

It means that for instance we could have

a b c: d e f

in a .am file, and have file_contents_with_transform (provided we
taught it that a, b and c are special targets) push @a, @b, @c the
list of d, e and f.

This would considerably improve the modularity IMHO.  There are many
dependencies pushed by hand in automake.in that really belongs to the
corresponding .am file.

Now my patch is only an approximation of this, since it catches
specifically just .PHONY.  How to extend it?

I suppose we should have file_contents_with_transform recognize
dependency lines without action, and process these lists, and remove
them from the actual contents.  I.e., special casing as is done in
this patch seems a dumb approach.

Still, we could apply it since it starts cleaning up some areas.

| Please add a ChangeLog entry for handle_tests.  Also, the ChangeLog

Groumph.  I remember I forgot my `prcs checkin' at home, and had a bad
time trying to separate these patch from the next ones.  And I lost my
ChangeLogs in the meantime.  Sorry I didn't pay more attention
restoring them.


| Finally, could you move the `Solaris 2.7' comment from handle_tests
| into check.am and make it into a `##' comment?  In fact, move it to
| just above the line it comments.

That's what was done :) :)


| It is fine to `cvs add' a file so that `cvs diff -N' creates the
| correct diff.  This applies generally -- if you don't have cvs write
| access there is a script you can get that will do a phony `cvs add' by
| manipulating CVS/Entries.

Oh, I didn't know that, thanks.  So why is the server contacted when
you cvs add?  To check that no one created one I guess, but what's the
point?  After all when two files are added at the same time, the diff
between the two is as well defined as the diff between two
`generations' of the `same' file.