[lfs-dev] LFS SVN urrent build results
g.esp at free.fr
g.esp at free.fr
Thu Dec 27 23:49:07 PST 2012
----- Mail original -----
> De: "Bruce Dubbs" <bruce.dubbs at gmail.com>
> À: "LFS Developers Mailinglist" <lfs-dev at linuxfromscratch.org>
> Envoyé: Vendredi 28 Décembre 2012 01:03:32
> Objet: Re: [lfs-dev] LFS SVN urrent build results
> Pierre Labastie wrote:
> > Le 25/12/2012 00:41, Bruce Dubbs a écrit :
> >> I'm not quite ready to give up on E2fsprogs-1.42.6 because I
> >> think it
> >> only is a problem in jhalfs. AFAICT right now, the test actually
> >> passes, but leaves an unwanted background process running. I'm
> >> not
> >> sure we want the -k in the make check. I get:
> >> 125 tests succeeded 0 tests failed
> >> Perhaps if we kill the debug fs in jhalfs when we finish the
> >> checks,
> >> that would clean things up.
> > I think I have found where the problem is:
> > from the tests log:
> > [...]
> > ../../tests/f_mmp/script: line 25: ps: command not found
> > f_mmp: disable MMP with tune2fs after e2fsck killed: ok
> > [...]
> > And ps is not found because procps is built after e2fsprogs...
> Yes. Found that yesterday and moving procps fixes the problem. The
> test (and the gzip problem I had earlier) are both OK now.
> > If I understand correctly the script,
> > ps is used to find debugfs PID, in order to kill it.
> > If the PID is not found, debugfs is not killed.
> > The funny thing is that if debugfs is not killed, the test should
> > not pass.
> > Maybe a bug report to upstream?
> I don't think so. This is quite specific to LFS.
The test try to check a feature of debugfs.
I agree that test should fail if PID is not found. But that's understandable
> > Anyway, procps should be added as a dependency of e2fsprogs,
> > and built before.
> > This should not be a problem since according to the `dependencies'
> > chapter, all procps's dependencies are built in chapter 5.
> I'll fix it either tonight or tomorrow.
> -- Bruce
I checked using procps-ng-3.3.4 build earlier that procps-ng test suite still pass.
I just had to adjust the hack that remain on the test suite due to 'which' not build
-sed -i 's|exec which sleep|exec echo /bin/sleep|' testsuite/config/unix.exp
+sed -i 's|exec which sleep|exec echo /tools/bin/sleep|' testsuite/config/unix.exp
Anyway, it should be doable to remove ps dependency from e2fsprogs test suite.
The idea is to save the PID during execution using echo $$ >PID and later $(cat PID) instead of using ps to find the pid.
First test I made using that idea didn't work, I need more testing.
More information about the lfs-dev