[Swig-devel] CCache test failures on Travis?

[Vadim Zeitlin]

Hello,

Does anybody know what is going on with CCache tests on Travis? They fail
for my PR 388 although I haven't absolutely anything related to CCache, see
e.g. https://travis-ci.org/swig/swig/jobs/59592274

Is it just a fluke or could I have really broken this somehow?

Thanks,
VZ

[William S Fulton]

Post by Vadim Zeitlin
Hello,
Does anybody know what is going on with CCache tests on Travis? They fail
for my PR 388 although I haven't absolutely anything related to CCache, see
e.g. https://travis-ci.org/swig/swig/jobs/59592274
Is it just a fluke or could I have really broken this somehow?

I couldn't see any differences in the Travis setup when comparing the
logs of the last good build and yours and I don't see anything in your
patch that would affect ccache. Strange. ccache fails the 1st test, so
its as if something fundamental has gone wrong. I've just rerun
#1364.2 https://travis-ci.org/swig/swig/jobs/58936952 which passed 5
days ago and now it fails. So either a lurking bug has been exposed
with some environment change or Travis is broken. I can't look right
now, but if anyone else has ideas, let us know.

William

[Vadim Zeitlin]

On Wed, 22 Apr 2015 23:02:03 +0100 William S Fulton <***@fultondesigns.co.uk> wrote:

WSF> On 22 April 2015 at 18:50, Vadim Zeitlin <vz-***@zeitlins.org> wrote:
WSF> > Hello,
WSF> >
WSF> > Does anybody know what is going on with CCache tests on Travis? They fail
WSF> > for my PR 388 although I haven't absolutely anything related to CCache, see
WSF> > e.g. https://travis-ci.org/swig/swig/jobs/59592274
WSF> >
WSF> > Is it just a fluke or could I have really broken this somehow?
WSF> >
WSF> I couldn't see any differences in the Travis setup when comparing the
WSF> logs of the last good build and yours and I don't see anything in your
WSF> patch that would affect ccache. Strange. ccache fails the 1st test, so
WSF> its as if something fundamental has gone wrong. I've just rerun
WSF> #1364.2 https://travis-ci.org/swig/swig/jobs/58936952 which passed 5
WSF> days ago and now it fails. So either a lurking bug has been exposed
WSF> with some environment change or Travis is broken. I can't look right
WSF> now, but if anyone else has ideas, let us know.

I wonder how much sense does it make to run CCache tests on Travis anyhow,
knowing that Travis reinitializes the build environment completely every
time making CCache useless. So perhaps we should just disable these tests?
VZ

[William S Fulton]

Post by Vadim Zeitlin
I wonder how much sense does it make to run CCache tests on Travis anyhow,
knowing that Travis reinitializes the build environment completely every
time making CCache useless. So perhaps we should just disable these tests?

Possibly you have misunderstood these tests. These tests are testing
the ccache-swig executable which we builid/distribute/install
alongside the swig executable. We absolutely should be testing any
binaries we are responsible for. The tests test all the features of
ccache-swig with a C compiler and SWIG. It really doesn't matter if
the build environment is fresh or not, the tests carefully control and
monitor the cache.

William

[Vadim Zeitlin]

On Thu, 23 Apr 2015 07:27:53 +0100 William S Fulton <***@fultondesigns.co.uk> wrote:

WSF> On 23 April 2015 at 00:17, Vadim Zeitlin <vz-***@zeitlins.org> wrote:
WSF> > I wonder how much sense does it make to run CCache tests on Travis anyhow,
WSF> > knowing that Travis reinitializes the build environment completely every
WSF> > time making CCache useless. So perhaps we should just disable these tests?
WSF>
WSF> Possibly you have misunderstood these tests. These tests are testing
WSF> the ccache-swig executable which we builid/distribute/install
WSF> alongside the swig executable.

Yes, I did misunderstand this, sorry -- and thanks for the explanation
(and separate thanks for fixing the build!).

VZ

[William S Fulton]

Post by William S Fulton

Post by Vadim Zeitlin
Hello,
Does anybody know what is going on with CCache tests on Travis? They fail
for my PR 388 although I haven't absolutely anything related to CCache, see
e.g. https://travis-ci.org/swig/swig/jobs/59592274
Is it just a fluke or could I have really broken this somehow?

I couldn't see any differences in the Travis setup when comparing the
logs of the last good build and yours and I don't see anything in your
patch that would affect ccache. Strange. ccache fails the 1st test, so
its as if something fundamental has gone wrong. I've just rerun
#1364.2 https://travis-ci.org/swig/swig/jobs/58936952 which passed 5
days ago and now it fails. So either a lurking bug has been exposed
with some environment change or Travis is broken.

CCACHE_DISABLE is now set in the Travis environment. The ccache tests
assumed none of the CCACHE_xxxx environment variables were set. I've
reset them all at the beginning of the ccache tests, so Travis testing
should work again now.

William