Re: [Boost-testing] [graph] betweenness_centrality_test takes more than 14 minutes

From: Martin Wille (mw8329_at_[hidden])
Date: 2005-04-14 02:09:29

Victor A. Wagner Jr. wrote:
> At Wednesday 2005-04-13 20:37, you wrote:

>> We need some way of profiling the tests as long as we don't have a way 
>> of limiting the execution time for them. Apparently, a couple of the 
>> tests take quite long to run and I only can find them out by staring 
>> at the screen while the tests are running. While excessive time 
>> consumption at first only annoys the testers, it also may indicate a 
>> flaw in the code. The amount of time (and RAM) consumed actually is 
>> important information.
> 
> 
> Martin, it _should_ be rather easy to add a timestamp to the test logs 
> each time a test is run.
> I wrote a very small application using boost::date-time which produces:
> C:\Documents and Settings\vawjr>timestamp
> 20050414T051138
> 
> adding the actual code (basically format the current time) should be 
> trivial for any booster

I was more concerned about how the postprocessing of the additional 
output would work than about how to create the timestamp itself. (The 
timestamps should be added by the build/test systems, I think)

process_jam_log would have to be adapted for that change.

I think it would be a good idea to allow the tests and the test system 
to spit out additional XML which would simply get copied into the XML 
files created by process_jam_log. Once that is done, adding new 
information should be a lot easier.

> as for the memory usage, could be tricky on windows.

If we add the additional-XML feature then we could make this step optional.

Regards,
m