From: cgf AT cygnus DOT com (Christopher Faylor)
Subject: Re: gcc --print-file-name: win32 or posix?
29 Dec 1998 18:27:49 -0800
Message-ID: <19981229210220.D9716.cygnus.cygwin32.developers@cygnus.com>
References: <19981229153224 DOT 54748 AT cygnus DOT com>
Mime-Version: 1.0
Content-Type: text/plain; charset=us-ascii
To: Mumit Khan <khan AT xraylith DOT wisc DOT edu>, cygwin32-developers AT cygnus DOT com

On Tue, Dec 29, 1998 at 06:29:37PM -0600, Mumit Khan wrote:
>Geoffrey Noer <noer AT cygnus DOT com> writes:
>> FYI, I've submitted the following patch to the EGCS folks.  Now it
>> continues to print Win32 paths but with forward slashes instead of
>> backslashes.  This should make everything happy (configure and make
>> seem to like the change at least...).
>
>As long as we understand the caveat that GCC now will not accept 
>paths like '-Ifoo\\bar' since most of the pathname logic uses the
>following type of code:
>  
>  if (path[x] == '/' || path[x] == DIR_SEPARATOR)
>   ...
>
>I of course am quite happy to see this change, but thought I should
>point this out. In fact, now that I think about, this is possibly
>the better compromise than what I wanted to do -- completely removing 
>support for non-POSIX paths (which included changing DIR_SEPARATOR
>to '/' and also removing the definition for GET_ENV_PATH_LIST). 
>
>With this change, the only possible complaint could be from folks
>using the output of `gcc -print-search-dirs', expecting ':' as the 
>path separator and getting ';' instead.

I am a little worried about that.  I am also worried about the fact that
since this bypasses the mount table a -print-search-dirs will output
absolute Win32 paths that could change tomorrow if someone decides that
they want to make a mount point mean something different.

I guess we'll see...

cgf