Re: parallel testsuite execution

[Akim Demaille]

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

 > Hello Benoit,

Hi Ralf,

Sorry for answering so late...


 > Can you rid it of nonportable constructs, and more importantly, can it
 > be included in Automake (copyright transfered to the FSF etc.)?

Sure, no problems.

 > Hmm, let's see.  Most suffix rules should be trival to convert, unless I
 > missed something, except for the

 >> %.log: %$(EXEEXT)

 > rule.

Yes, it should be straightforward.  In fact the most difficult task is
that I'd like to support multiple test suites in a single directory.
Maybe that's overkill, but in my context it would be quite useful
actually.

So I would like to be able to write something like

TEST_SUITES = foo bar
foo_TESTS = foo1.chk foo2.test foo3
bar_TESTS = bar1.test bar2.sh bar3.c

and be able to write on the side rules explaining how to .chk -> .log,
.test -> .log etc.

Maybe instead of _TESTS we could reuse _SOURCES?  That's not
inconsistent: they are really sources which must be compiled into
*.log files which are then linked together to produce foo.log and
bar.log.

 > I guess as a first approximation it would be ok to do without.  The
 > awk script may require a bit of work for Solaris; I'd just drop the
 > colors;

Nah, please, let's keep them, it's really very useful!

 > $(basename ..) is not portable;

Will be fixed when migrated to Automake.

 > and also it would need adjustment for Sun make's gigantic VPATH
 > rewriting feature; 'TEST_LOGS ?=' would need to be replaced by an
 > override done at 'automake' time.  

Sure, but who's going to spend some time on this?

 > The whole thing should probably be governed by an Automake option
 > parallel-tests, defaulting to off, for backward compatibility.
 > Then some documentation, and about three tests to ensure it works
 > as intended.

We can handle that part.

 > IOW, if nobody else I could probably even volunteer to do the rewrite
 > once the legal situation is clear.

That's a good thing to know :)