|
|
Boost Testing : |
From: Markus Schöpflin (markus.schoepflin_at_[hidden])
Date: 2005-05-04 09:02:36
Douglas Gregor wrote:
>
> On May 4, 2005, at 8:13 AM, Markus Schöpflin wrote:
>
>> The tests are run in an endless loop (as long as the machine is not
>> restarted) and the expected turnaround for a full regression build is
>> something around 4 days, so please be patient when you want to check
>> something. :-)
>
> Are these tests run on the same machine as the tru64cxx65-042 tests?
Yes, they are.
> If so, it might actually be more productive to only run the
> tru64cxx65-042 tests (which would cycle in 2 days or so, I'd guess?).
Yes, it takes about two days for a full cycle with that compiler.
> Coverage is important, but without significant throughput we won't be
> able to address problems on that platform and still keep a timely
> release schedule.
I added the tests because I think it might help sort out which issues are
platform issues and which issues are compiler related. I'm hoping that the
incremental runs will have a much shorter turnaround time than the full
runs. So maybe this is not so much of a problem.
I'll keep the tests running for a week or so and if the turnaround really
is too slow, I'll remove the gcc tests again.
The biggest problem currently is the wave library which seems to hang cxx
on that platform. I had a successful regression run for it last Friday, but
when I check out with that timestamp, the compiler hangs anyway. So maybe
I'm just not patient enough...
Markus