From: Markus Schöpflin (markus.schoepflin_at_[hidden])
Date: 2002-10-01 07:34:10
David Abrahams wrote: > From: "Markus Schöpflin"
<markus.schoepflin_at_[hidden]>
>
>> Hmm, as you asked I investigated a little further. Seems like the
>> vacpp toolset needs another fix. To quote from the documentation
>> for vacpp:
>>
>> <quote> The default option, -bdynamic, ensures that the C library
>> (lib.c) links dynamically. To avoid possible problems with
>> unresolved linker errors when linking the C library, you must add
>> the -bdynamic option to the end of any compilation sections that
>> use the -bstatic option. </quote>
>>
>> This looks like we could unconditionally add -bdynamic to the end
>> of the xlc-Link-action. If we decide to do this, I will do a clean
>> regression run before commiting that fix.
>
> Sounds good to me.
Ok, I changed the vacpp toolset definition file and the regression run
is ok. I will commit the fix to the release branch.
>> But I still think that the Jamfile should'nt globally force
>> requirements just because a specific compiler needs them.
>
> Given the nature of the problem it seemed likely to me that more
> compilers would have the same requirement. For details, ask the
> Metrowerks guys about why there's no locale support in their dynamic
> library implementation.
>
> Anyway, I think it would be a bad idea to change it from static to
> dynamic on the release branch at this late date. Who knows what could
> break?
The worst that could break is the io state saver test because the
problematic line is in the Jamfile in the status directory. :-)
Anyway, I have no problems if you prefer not to change it as the test
passes with the fixed toolset defintion.
Markus