|
|
Boost Testing : |
From: Robert Ramey (ramey_at_[hidden])
Date: 2005-07-18 10:18:34
I'm very pleased with the TRU64 test results. I presume that no special
command switches were required to it looks like we're done.
Except for the variant test. The variant test seems pretty simple to me.
Your comment suggests that there is some sort of code violation either in
boost/serialization/variant.hpp or in the test. I don't see it. Do you
have any idea where it is?
Robert Ramey
Markus Schöpflin wrote:
> Robert Ramey wrote:
>
>> Well, for now, I would be happy to add the necessary switches to the
>> serialization test Jamfile and then add a note to the documentation.
>
> The problem is, I have no idea how to do this right with boost.build
> short of adding explicit support for this in the build system.
>
> Currently, -timplicit_local gets added by the toolset whenever a C++
> program file is compiled. You would have to _remove_ this option from
> the compiler command line and I don't think that is possible.
>
>> I am puzzled by one thing. I was under theimpression you had a
>> TRU64 system at your disposal and were using it to track down this
>> problem. Is there a reason you want to wait for another round of
>> tests before adding the magic switch to to the serialization test
>> Jamfile?
>
> Two reasons: As outlined above it's not a question about adding a
> specific command line switch but about removing one. And secondly, I
> was hoping that it would work anyway because of your usage pattern of
> the statics. And as the latest tests results show, only the variant
> tests are failing now, and this is because variant is broken with
> this compiler in strict_ansi mode.
>
> Markus