Acceptable outcomes of SwarmSolverTest::testUnconstrained
quikee at gmail.com
Wed Feb 28 09:59:35 UTC 2018
On Wed, Feb 28, 2018 at 5:41 PM, Stephan Bergmann <sbergman at redhat.com> wrote:
> Those ASan+UBSan tinderbox builds execute rather slowly, yes.
> (<http://clang.llvm.org/docs/AddressSanitizer.html> claims "Typical slowdown
> introduced by AddressSanitizer is 2x.")
> But also as reported by others today on #libreoffice-dev:
>> Feb 28 09:17:32 <buovjaga> sberg: I got a swamsolver failure
>> yesterday. Then I pulled later and the next build went fine.
>> Feb 28 09:19:03 <buovjaga> After the failure, soffice refused to
>> start. I don't have logs, unfortunately
Hmm.. if only testUnconstrained is the problem I can live with it
being disabled (or just not check the result). For this one we expect
that the algorithm finds a solution without any constraints, which
means the initial values are generated randomly in a range from -inf
to inf (min float to max float actually), then expect that they
converge towards the solution, which can only happen consistently in a
large number of generations. Other tests have sensible constraints so
they find the solution much quicker and are more important to me.
Running solver without constraints is discouraged anyway.
So the patch in  should disable the test.
More information about the LibreOffice