Re: 39-new-test-suite

[Alexandre Oliva <oliva at lsd dot ic dot unicamp dot br>]

[François, please find a question for you in the last paragraph]

On Dec 22, 1999, Akim Demaille <akim@epita.fr> wrote:

> When it comes to handle this one, I'd appreciate if someone could tell
> me how I should handle renaming/moving files with CVS.  Should I
> remove+add?

See my reply to 40-*

> +	Introduce a new style of testing, based on the framework
> +	developped by François Pinard in Free Recode 3.5.

> +	Introduce the logistics.

Ok

> +	Write tests.

Ok

> +	The test syntax.m4 tries to run all the specific macros, but
> +	hasbeen'd macro make configure die with bad exit status.  So we
> +	wish to distinguish live macros from defunct macros at the moment
> +	they are AC_DEFUN'd, not in their body.  Hm, defunct,
> +	defunct... Hey!  That's a good name!

:-D :-D

Ok

> +	* autoconf.sh: Use cat -s to nuke multiple empty lines.

Not portable.  Solaris' cat, for example, says:

     -s    cat is silent about non-existent files.

Ok, it wouldn't hurt, but what if some other vendor decided to give
`cat -s' some weird meaning?  It should be easy enough to discard and
not count duplicate blank lines within the awk script.

> +## There is currently no means with Automake not to run aclocal.
> +ACLOCAL_AMFLAGS = --version >/dev/null && touch aclocal.m4
 
BTW, it would probably be nice to arrange that configure depends on
the m4 files.  Eventually, automake will do that, but, meanwhile,
let's do it manually.

> +dnl AC_DEFUNCT(NAME, COMMENT)
> +dnl -------------------------
> +dnl Declare the macro NAME no longer exists, and must not be used.
> +define([AC_DEFUNCT],
> +[define([$1], [AC_FATAL([$1] is defunct[$2])])])

DEFUNCT is nice for a macro name, but I don't think it's nice for a
message for users/developers.  How about s/defunct/no longer
available/?

> Index: 0.62/autoheader.m4
> --- 0.62/autoheader.m4 Mon, 01 Nov 1999 05:29:28 +0100 akim (ace/24_autoheader 1.18 664)
> +++ 0.62(w)/autoheader.m4 Thu, 11 Nov 1999 15:35:24 +0100 akim (ace/24_autoheader 1.18 664)
> @@ -79,8 +79,8 @@
>  dnl the newline, which makes the @@@ not always be at the beginning of
>  dnl a line.
 
> -dnl AH_DEFINE(VARIABLE [, VALUE[, DESCRIPTION]])
> -dnl --------------------------------------------
> +dnl AH_DEFINE(VARIABLE, [VALUE], [DESCRIPTION])
> +dnl -------------------------------------------
>  dnl When running autoheader, this macro replaces AC_DEFINE and
>  dnl AC_DEFINE_UNQUOTED.
>  dnl

Separate check-in.  Being just a cosmetic change, no ChangeLog entry
is required (and, in fact, there is none), but checking it in with the
same message used for the other changes would be confusing at best.

> +AC_INCLUDES((m4/atconfig.m4, m4/init.m4, m4/missing.m4, m4/sanity.m4))dnl

I'm still not sure I like this AC_INCLUDES.  Its view doesn't make it
any clearer than multiple AC_INCLUDE statements to me.  But this is
just personal opinion.

> +AT_CONFIG(..)
 
A comment in this line explaining that it sets up the testsuite would
be welcome :-)

> +##m4-files-begin
> +EXTRA_DIST = README Makefile.am.in \
> +atconfig.m4 \
> +init.m4 \
> +missing.m4 \
> +sanity.m4
> +##m4-files-end

Homework assignment: write a Makefile.am.in that outputs itself :-)


How about generating the list of m4 dependencies in a separate file,
as a single Makefile macro definition, with $(top_srcdir)/ before each
filename, and use automake's `include' feature wherever the list of
files might be used?  For example, as a dependency of `Makefile.in's,
configure, and even aclocal.m4?  Then we wouldn't need Makefile.am.in.

> +	m4 -I $(srcdir) atgeneral.m4 suite.m4 | cat -s > $@-tmp
                                                    ^^ non-portable

> Index: 0.62/tests/suite.m4

> +# Validation suite for the Free `recode' program and library.

Oops :-)  You may want to change this.

> +# Copyright © 1998, 1999 Progiciels Bourbeau-Pinard inc.
> +# François Pinard <pinard@iro.umontreal.ca>, 1998.

Shouldn't this be copyrighted by the FSF?  I see François has
assign.future in file for autoconf, but we can't just ``steal'' this
file, nor can we ship autoconf with it without a FSF copyright.  I
think :-).  François?

-- 
Alexandre Oliva http://www.ic.unicamp.br/~oliva IC-Unicamp, Bra[sz]il
oliva@{lsd.ic.unicamp.br,guarana.{org,com}} aoliva@{acm,computer}.org
oliva@{gnu.org,kaffe.org,{egcs,sourceware}.cygnus.com,samba.org}
** I may forward mail about projects to mailing lists; please use them