GCC 2.95.3-1 available for testing -- question for Earnie

Christopher Faylor cgf@redhat.com
Sat Mar 24 20:24:00 GMT 2001


On Sun, Mar 25, 2001 at 04:10:35PM +1200, Danny Smith wrote:
>--- Christopher Faylor <cgf@redhat.com> wrote: > On Sun, Mar 25, 2001 at
>02:21:07PM +1200, Danny Smith wrote:
>> >> Sorry.  It searches /usr/includw/mingw/g++-3 explicitly.  Doesn't it
>> >> make sense to put the mingw headers there?  Are the headers generic
>> >> enough that this doesn't matter?  
>> >
>> >The g++ headers in cygwin are *identical* to the ones that gcc make install
>> put
>> >in my mingw package.
>> 
>> How have you verified that?  Did you actually diff the directories?
>> 
>I have verified this with for 2.95.3. diff -urpb produces this:
>
>"Only in /d/mingw/include/g++-3: dummy.h"
>where dummy.h is a empty file I put into to verify that diff was working.
>
>But....
>changes involving _IO_fpos_t/_IO_off_t in libio.h (and other public headers) in
>gcc-2.95.3 can cause backward compatability problems (with mingw, not cygwin)
>if we patch the definitions in libio  .cc src files to match the new
>prototypes, which is what it appears that you have done.  Eg.
>istream::seekg(long long) in mingw 2.95.2 has become istream::seekg(long) in
>2.95.3 beacuse of changes in typedef of streampos from __IO_fpos_t to
>__IO_off_t. This problem only arises if we link objects compiled with 2.95.2
>against the 2.95.3 libstdc++ .  I repeat this does not affect cygwin, only the
>__MSVCRT__ version of mingw. 
>
>If -mno-cygwin  wants to retain backward compatabilty (at least until G++-v3
>breaks it in a big way), perhaps mingw headers should be different (ie, revert
>the type changes in libio.h).

AFAIK, I just applied gcc patches that were part of the cygwin net
release to gcc 2.95.3.  Any changes that you see are part of the
standard 2.95.3.

This is YA reason for giving mingw its own include area.

Or, maybe it's another reason for eliminating -mno-cygwin entirely and just going
with a separate i686-pc-mingw-gcc compiler instead.

cgf

--
Want to unsubscribe from this list?
Check out: http://cygwin.com/ml/#unsubscribe-simple



More information about the Cygwin mailing list