This is the mail archive of the
gcc@gcc.gnu.org
mailing list for the GCC project.
RE: Notes from the testing BOF at the summit
- From: "Dave Korn" <dk at artimi dot com>
- To: "'Ben Elliston'" <bje at au1 dot ibm dot com>
- Cc: "'Alexandre Oliva'" <aoliva at redhat dot com>,<gcc at gcc dot gnu dot org>
- Date: Wed, 9 Jun 2004 11:07:28 +0100
- Subject: RE: Notes from the testing BOF at the summit
> -----Original Message-----
> From: Ben Elliston
> Sent: 09 June 2004 06:35
> To: Dave Korn
> Cc: 'Alexandre Oliva'; gcc
> Subject: Re: Notes from the testing BOF at the summit
>
> "Dave Korn" writes:
>
> > > Sorry, I wasn't very clear. Someone pointed out that
> DejaGnu needs
> > > parallelism that extends down to individual files within
> a single .exp
> > > test script.
>
> > IIUIC that's exactly what you have there: notice the command line
> > specifies both the .exp file and the individual test within the .exp
> > file (that's what the "=testname" bit is about). You can also use
> > pattern matching.
>
> I don't think you're following me .. :-) We're talking about _parallel
> testing_.
>
> Ben
Well, if you use parallelism in the makefile and use that command line
syntax to subdivide the tests into a bunch of subsets that can be run in
parallel.... doesn't that basically give you what you want?
Or if you're referring to running a single test with all the different
sets of compiler flags in parallel, can't you just set up a bunch of
parallel make check targets each restricted to one of the different option
sets?
check:
$(MAKE) -j5 check-O0 check-O1 check-O2 check-O3 check-O3-g [...etc...]
check-O0:
runtest RUNTESTFLAGS=--target_board "unix{-O0}"
check-O1:
runtest RUNTESTFLAGS=--target_board "unix{-O1}"
[...]
check-O3-g:
runtest RUNTESTFLAGS=--target_board "unix{-O3 -g}"
[...etc...]
Basically what I'm saying is that, given the ability we already have to
filter and restrict the test set on any given invocation, isn't using
makefile parallelism enough to do the rest?
cheers,
DaveK
--
Can't think of a witty .sigline today....