This is the mail archive of the
libstdc++@gcc.gnu.org
mailing list for the libstdc++ project.
Re: [PATCH, libstdc++]: Increase timeout for testsuite/ext/pb_ds/regression tests
On Tue, Jan 13, 2009 at 1:03 PM, Paolo Carlini <paolo.carlini@oracle.com> wrote:
>> Attached patch increases timeout for some long tests in
>> testsuite/ext/pb_ds/regression directory by a factor of 2.0. This
>> fixes all timeouts for alphaev56-unknown-linux-gnu cfarm machine.
> This patch seems so simple and safe (and it is! ;) but I'm bit concerned
> by the explosion of different strategies we are using to deal with
> timeouts and long runtimes for special targets. For example, can't we
> unify the way we deal with simulators and alpha*?
I think that simulators represent different class of targets, so they
should be treated separately. In fact, the test is changed for the
simulator by limiting the number of cycles, and IMO this is not
desirable for native targets i.e. - alpha* would run a different test
than other targets.
Also, recent usage of dg-timeout-factor has shown, that adding target
selector to this dg-directive is more of an exception than a general
rule. If the test timeouts on one target, it is quite possible that it
will timeout on others too.
Uros.