Re: [boost] Regression Tests, Boost.Serialization

From: Martin Wille (mw8329_at_[hidden])
Date: 2004-07-16 02:14:25

Robert Ramey wrote:
> Jeff Garland wrote:
> 
> 
>>>Houston, we have a problem.
>>>
> 
> 
>>>I had a few GB of disk space left a few days ago.
>>>Now, I'm unable to run all the tests. I could
>>>drop one of the gcc-3.4 versions. However, I'd
>>>still be short by ~10GB after that.
> 
> 
> It's a fact - running the bjam serialization test is going to consume 20 GB.

The disk requirements can be changed. Stripping the test binaries
and removing the .o files already has been mentioned. Another
chance lies in merging similar (in terms of instantiated templates)
tests into one binary.

> If the test machine doesn't have it the best is just to remove the
> serialization test from the status jamfile for that machine.

That'd be quite unfortunate an solution.

   Actually I
> think it would be unusual for a machine used for testing boost not to have
> this much free space available. So I would hope it only affects martin's
> tests.

Well, I don't think I have to expect the disk requirements to
be tripled just because one library is being added to Boost
and to save disk space just for that case. Not having some
large company-owned box around for testing, I need to plan
the resources. I wouldn't be surprised if other volunteers
were in a similar situtation.

>>It would be nice if we could drop serialization on compilers that just
>>aren't going to work. 
> 
> 
> Having looked at the test results on all the compilers the only ones that I
> really wonder about are HP, and for now Comeau.  All tests run with MSVC 6.0
> and borland will fail without spirit 1.61 being installed.  This is because
> spirit 1.61 is needed to build the libraries.  So if the library build fails
> then all the tests are skipped.  They are marked "fail" which is a little
> misleading - "skipped" might be better.

Ah, good! This explains why gcc-2.95 doesn't need that much space
for the tests :)

[...]
>>As for what we should do now, it would be nice to have an answer to the
>>question I posed this morning: is serialization going into 1.32 or are we
>>cutting it out in the final release:
> 
> 
> I want to keep working to finish regardless of which release it becomes a
> part of.
> 
> 
>>If we are cutting it out the we can just remove it from status bjam now....
> 
> 
> But that would mean that progress would be stopped on getting this done.

Right. I do not ask to remove Boost.Serialization from the tests.
That'd be the sadest thing we could do.

Regards,
m