re: (patch) Workaround for gcc's float complex bug.

[Mumit Khan <khan at NanoTech dot Wisc dot EDU>]

On Mon, 22 Nov 1999, Benjamin Kosnik wrote:

> 
> Hey Mumit, thanks for this work. It's in, with some minor mods. All the
> macros have to have _GLIBCPP_ prepended to them, so:
> 
>         * acconfig.h: New _GLIBCPP_BUGGY_FLOAT_COMPLEX macro.
>         * src/complexf.cc: New _GLIBCPP_FLOAT_SPECIALIZATION macro.
> 
> I just went ahead and did these changes. 

And I actually took out the _GLIBCPP prefix before sending it off!
I'm trying to get these changes out of my local tree before I start
forgetting about them. Thanks for dealing with these so quickly.

> Does this mean that I can compile v-3 under cygwin-release-one now? I
> tried a week or so ago and had crashers in make (perhaps related to
> this?) I would be interested in getting CVS libstdc++ to compile under
> cygwin, so I look forward to your other patches.

There are two more tiny patchlets - 
  - Cygwin doesn't have WEOF. I have provided a workaround, which is
    certainly not correct for the long term.
  - workaround for a crash in the compiler. I already submitted a patch 
    for bits/locale_facets.tcc for this, but missed a few places somehow.

I'll send both off today. 

The big problem with Cygwin build is that the DLL gets hosed when running 
the gen-limits.exe; I'm using a patch from Vadim Egorov[1], not (yet) 
sanctioned by Cygwin team, that allows me to build it. What I can do is 
to upload the bits/std_limits.h and src/limitsMEMBERS.cc to my web page 
for those who want to build it under Cygwin. I'll write up a FAQ for 
building on Cygwin and that may help others doing the same.

> As an aside, the continuing devolution of the math/ctype headers is
> bumping up the priority of getting a clean fixincludes-ish solution in
> to automatically prune headers, so that the standard headers are not
> littered with macros. . .

I agree completely. It's going to take some work however. The first time 
I tackled the Mingw32 macros, I tried it, and gave up quickly after I saw
the nesting of the macros (which is going to be some work in unravelling 
now).

I'll send off the Cygwin bits this evening after I merge it into updated
CVS.

Regards,
Mumit

[1] For Vadim's "setjmp/longjmp patch update" for Cygwin, see
http://sourceware.cygnus.com/ml/cygwin-developers/1999-09/msg00006.html