Re: [Condor-devel] trouble with condor_fix_assert and undef __GNUC__

[Dan Bradley <dan@xxxxxxxxxxxx>]

I think this can be easily resolved, because I was overlooking an 
important detail.  Near the end of this file, there is some code that is 
supposed to restore __GNUC__ to its original value.  However, this was 
not restoring it in the case where __GNUC__ was originally 4.  It only 
restored for GNUC 2 and 3.

I still don't see why we are grouping together all non-linux platforms 
when doing this, but at least it does not have as far reaching 
consequences (when correctly done) as I originally thought.

--Dan

Dan Bradley wrote:
> In condor_fix_assert, it undefines __GNUC__ on all platforms except for 
> LINUX.  What do all non-LINUX platforms have in common with respect to 
> the use of __GNUC__ in header files?  That seems like a bad way to group 
> things.  I bring it up because it is conflicting with the classads 
> external when I try to compile condor with it on Mac OS X.
>
> The supposed reason for doing it is, "Get rid of __GNUC__, otherwise gcc 
> will assume its own library when
> generating code for asserts."  If this undef is not needed on Linux, why 
> should we have it anywhere else?    I propose that we get rid of this 
> undef __GNUC__ altogether.
>
> --Dan
>
> _______________________________________________
> Condor-devel mailing list
> Condor-devel@xxxxxxxxxxx
> https://lists.cs.wisc.edu/mailman/listinfo/condor-devel
>