On Wed, Feb 11, 2015 at 11:01 PM, Branko Čibej <brane_at_wandisco.com> wrote:
> On 11.02.2015 20:07, Branko Čibej wrote:
> > On 11.02.2015 20:03, Philip Martin wrote:
> >> Branko Čibej <brane_at_wandisco.com> writes:
> >>> I'm seeing this in the logs on the svn-x64-macosx-bdb builder on trunk:
> >>> $ cat fails.log
> >>> [[[
> >>> XPASS: fs-test 44: test reopen and modify txn [[WIMP: txn_dir_cache
> fail in FSFS]]
> >>> ]]]
> >>> Philip, I think that's your pigeon; can you please check if this is a
> >>> fluke, or if we can remove the XFAIL/Work-In-Progress tags from this
> >>> test case?
> >> The test shows a bug in FSFS, on that backend the test will XFAIL.
> >> There is no corresponding bug in BDB of FSX so there the test is an
> >> XPASS. XPASS/WIMP appears to be the best way to describe a C test that
> >> behaves that way. I suppose XFAIL/WIMP might also work. I don't think
> >> there is an easy way to mark a C test as PASS on some backends and FAIL
> >> on others.
> > You're right, there's not. OK, as long as this is expected, fine.
> > Of course, we can always invent new predicate macros for the C tests ...
> > I'll see if I can come up with an FS-type-aware predicate.
> r1659101 introduces run-time predicates to the test suite
> infrastructure. We used to have only compile-time conditions, and that's
> clearly not enough to test the FS type, which is a run-time property.
> That test now is now marked as XFAIL, unless the FS type is not FSFS, in
> which case it's a PASS.
I was about trying to return SVN_ERR_TEST_SKIPPED
as a stop-gap. But you solution is clearly better.
Received on 2015-02-12 10:30:57 CET