Re: Regression when building static libraries with MSVC

Thore Karlsen <sid-m3MN3Z/L/[email protected]> Tue, 16 Aug 2005 22:49:32 -0500
Newsgroups gmane.comp.tools.aap.devel
Organization I've resigned.
Message-ID <[email protected]>
On Tue, 16 Aug 2005 09:22:17 -0500, Thore Karlsen
<sid-m3MN3Z/L/[email protected]> wrote:

>[... reason for C_BUILDLIB_ACTION/CXX_BUILDLIB_ACTION patch ...]

>>Ah, yes, that's where it came from.
>>
>>It's a bit of a puzzle, especially because I can't see how
>>C_BUILDLIB_ACTION would ever be used.  I think it isn't.  After all,
>>when building a library your input is object files, not C files.
>>
>>I think the original problem that Lars ran into is that the default
>>recipe uses this:
>>
>>    defer_var_names = ["BUILDLIB_ACTION", "{buildlibaction}"])
>>
>>This means that the BUILDLIB_ACTION variable is checked before the
>>buildlibaction attribute on the object files.  I think that's the wrong
>>way around: the attribute should be preferred over a global variable.
>>
>>What the change from Lars did was simply remove the BUILDLIB_ACTION, so
>>that the buildlibaction attribute would be used.
>>
>>I'll switch the order of the defer_var_names and undo the change that
>>Lars made.  Then it should be tested, because I'm not very sure this is
>>the final solution!

>Sounds good, I'll test this with MSVC. I don't have D installed, so I
>can't see if that is broken.

OK, I've tested 1.079, and my MSVC projects build just fine.

-- 
Be seeing you.



-------------------------------------------------------
SF.Net email is Sponsored by the Better Software Conference & EXPO
September 19-22, 2005 * San Francisco, CA * Development Lifecycle Practices
Agile & Plan-Driven Development * Managing Projects & Teams * Testing & QA
Security * Process Improvement & Measurement * http://www.sqe.com/bsce5sf