|
|
Boost Testing : |
From: Michael Glassford (glassfordm_at_[hidden])
Date: 2005-04-19 09:26:59
Martin Wille wrote:
> Paul Baxter wrote:
>
>>> Martin Wille wrote:
>>>
>>>> Hello,
>>>>
>>>> as Stephen (and others) reported 5 days ago, some of the mutex tests
>>>> take unacceptably long to run. (I'm watching an apparently deadlocked
>>>> read_write_mutex test "run" for several minutes right now; I don't
>>>> know when it started)
>>>>
>>>> Several operating systems are affected (at least Darwin, Solaris,
>>>> Linux).
>>>
>>>
>>>
>>> It's actually *a lot* worse than I first thought:
>>>
>>> I manually ran test_read_write_mutex (using the intel-8.1 version of it)
>>> and found out it takes 50 (yes, fifty!) minutes.
>>
>>
>>
>> Yep, I was about to report this too. I think it gets built/tested in a
>> couple of variants... very slow for a full test cycle
>
>
> D'oh! I'm running tests for seven toolsets. This test gets executed in
> two variants. The time taken for *all* tests is ~24h here. 2*7*50 min ==
> 700 min == 11.7h; that's almost 50%. This doesn't even take into account
> the excessive time consumption reported by Stephen for test_mutex.
>
I have some new tests that I'm working on that should be much faster.
The reason the old tests are so slow is that they use Sleep *a lot*. I
think I've come up with decent tests that hardly sleep at all;
hopefully I'll be able to check them in some time today.
Mike