Re: Partial shadow header patch (was Re: still problems withlibstdcxx-v3 and -ansi)

[Benjamin Kosnik <bkoz at redhat dot com>]

> > Any word on an ETA? Can you post a partial patch, or a work-in-progress?
> 
>   Okay, Okay already.  This should be considered work in progress.  Anyone
> whose interested (Brent?) might try it, but its definately not ready to go into
> the tree.

great! thanks.

>   Configure as normal for cshadow and add to your CXXFLAGS for any target code
> with "-fno-builtins -D_ISOC99_SOURCE -I $(prefix)/g++-v3/shadow".  For some
> stuff adding "-I $(prefix)/g++-v3/cshadow" maybe counterproductive.

A couple of comments.

1) according to the glibc manual,
1.3.4 Feature Test Macros
_GNU_SOURCE includes _ISOC99_SOURCE

since _GNU_SOURCE is already defined in the makefiles and in c++config.h,
does it (really) need to be defined on the command line?
(Ulrich, any comments?)

2) I'll add -fno-builtins to the makefiles, as per the email from earlier
today regarding built-in functions not being in the correct namespace.

>   There maybe some problems with the ordering of #includes in user code; I think
> mostly due to cshadow hacks or lack there of.  I wouldnt be surprised if
> non-glibc systems have problems, especially with the changes to _G_config.h.

I'm interested in getting this checked in so more people can play with it.

>   Oh, and this will work (most of the time) with -ansi -pedantic but only if
> you specify -D_ISOC99_SOURCE...

-ansi isn't a particular concern. 

-pedantic is....