On Thu, Mar 20, 2008 at 12:50:31PM +0000, Daniel P. Berrange wrote:
On Wed, Mar 19, 2008 at 03:27:30PM +0100, Jim Meyering wrote:
> A week or so ago, I read a bug report on this list with a simple
> reproducer + fix that was just begging for a test suite addition.
>
> This change adds framework to make that easy, and adds the first test.
> I uncovered problems in tests/Makefile.am along the way and just
> posted a patch for that.
>
> This adds most of a portable (bourne-shell-based) test framework that also
> provides for convenient isolation of individual tests (so we can easily
> parallelize them without worrying one will tromp on files of another).
> In a couple years when running "make check" runs hundreds of tests,
> you'll definitely want them to run in parallel. The mktempd script is
> not new. I first used it in coreutils and parted, but removed it from
> coreutils after that package acquired its own C-based implementation.
>
> test-lib.sh is from coreutils, but was inspired by the file by the
> same name in git.git's own t/ (tests) directory.
All seems reasonable to me.
For libvirt yes. For libxml2 on the opposite it proved useful to be able
to aggregate regression tests as a few runnable binaries, first this made
it like 10 times faster, but also allowed to run the regression tests on
plaforms (MSC, VMS ...) where we could not rely on make and unix shell.
For libvirt it's unclear the second point will ever apply, for the first
point the main elapsed time seems taken by make check on DNS checks.
But I don't have strong opinions one way or another honnestly, as long
as make check doesn't take 10mn and the output FAIL/PASS is made obvious
that's just fine by me.
Daniel
--
Red Hat Virtualization group
http://redhat.com/virtualization/
Daniel Veillard | virtualization library
http://libvirt.org/
veillard(a)redhat.com | libxml GNOME XML XSLT toolkit
http://xmlsoft.org/
http://veillard.com/ | Rpmfind RPM search engine
http://rpmfind.net/