gcc-2.95.3-4 to mingw

[Danny Smith]

--- Christopher Faylor <cgf@redhat.com> wrote: > Ok.  If I am going to
be applying these patches, then I have a few
> observations.
> 
> 1) They should not have MS-DOS line endings.
> 
> 2) They should not reference MS-DOS paths.
> 
> 3) I'm not going to add the word "mingw" to the compiler output for
> the cygwin
>    gcc compiler.  You are, of course, welcome to change that for any
> mingw
>    compiler.  I assume that you are doing that already.
> 
> 4) I'd appreciate a different ChangeLog entry for each directory that
> is
>    modified.  I can't use the one that you sent since it is
> non-standard.
> 
> I'll try to use the patch that was submitted here by stripping the
> CR's
> and changing all of the paths to '/' but I would appreciate it if
> future
> submissions adhered to the above criteria.
> 

Thank you.

> >One other little bit, a recurring problem with sys_nerr and
> >sys_errlist[] and dllimport in libiberty/strerror.c.  This should
> >probably be sent to libiberty; I haven't included in ChangeLog.
> 
> Also, it seems to me that this patch is being handled in the wrong
> way.
> There is already a configure test for this functionality.  Why is it
> being handled incorrectly in the mingw case?

Configure tests for libiberty say yes mingw does have strerror,
sys_nerr and sys_errlist.  The problem is libstdc++ which says that
liberty's strerror.o is needed (by iostrerror).  I'll have a look at
the configure process there (a problem because I can't seem to find any
config.log for libstdc++).   Your right--the fix should be in libstdc++
which  doesn't really need its own strerror, but could use the one in
the C-runtime.

Danny

> 
> cgf


_____________________________________________________________________________
http://messenger.yahoo.com.au - Yahoo! Messenger
- Voice chat, mail alerts, stock quotes and favourite news and lots more!