Re: Preprocessor problem on Cygwin

[Christopher Faylor <cgf at redhat dot com>]

On Tue, May 22, 2001 at 12:15:58PM +0200, Andreas Jaeger wrote:
>Alexandre Oliva <aoliva@redhat.com> writes:
>>Every now and then, I try my (bad) luck on Cygwin.  This time, I fell
>>prey of an apparent bug in the version of the Cygwin dll I had
>>installed: it sets errno even when lstat() succeeds, which breaks cpp's
>>pathname simplification code.
>>
>>Even though I believe this is actually a bug in Cygwin, I thought I'd
>>work around the problem, that might well be present in other OSs, with
>>this simple patch in GCC.  Ok for mainline?  Ok for branch?
>
>It's not a bug in Cygwin.  Errno is in general only defined when an
>error occurs, the library might change it if no error occurs.  The
>patch is therefore from a library perspective,

I agree.  Any software which relies on the value of errno to be
consistent across library calls or, worse, which thinks that it can get
valid information from errno after a *successful* call is IMO broken.

However, we will fix the problem in Cygwin since the setting of errno
was unintentional.  I believe that gcc should be fixed as well.

cgf