Re: g77's use of assert and x-compilation

[Ian Lance Taylor <ian at cygnus dot com>]

   Date: Wed, 22 Jul 1998 10:35:36 -0500 (CDT)
   From: Mumit Khan <khan@xraylith.wisc.edu>

   > For cygwin32, we just arranged to provide __eprintf in newlib, so that
   > all cases were covered.

   That of course covers all cases for *just* cygwin32.

Yes.  I was just pointing out another approach that can work.

   I build for two other
   win32 targets, mingw32 and uwin, that don't use newlib and incidentally
   don't have __eprintf tucked away somewhere else. 

Purely out of curiousity, do these targets have no supplied library
code at all?

   I still maintain that definining inhibit_libc in libgcc2.c is not right, 
   and it should be done via configure when cross-compiling *and* stdio.h is 
   not found.

I agree that the current code in libgcc2 is wrong.  As you may know,
the definition came over from gcc 2.8:

Wed Nov 26 05:05:36 1997  Richard Kenner  <kenner@vlsi1.ultra.nyu.edu>

        * libgcc2.c (inhibit_libc): Define #ifdef CROSS_COMPILE.

I believe that patch will be wrong more often than it is right.

In some sense the right thing to do is to run some autoconf tests
using the target compiler to see what is available.  Of course, that
can not be done until the target compiler is built.  That is an
extension of your suggestion of checking whether stdio.h is available;
how do you plan to write that check?

Ian