proposal: substitute-path handles foreign dir separators

[Pedro Alves]

On Friday 17 December 2010 19:43:56, Pedro Alves wrote:
> On Friday 17 December 2010 19:23:22, Eli Zaretskii wrote:
> > Thanks.  I have 2 comments:
> > 
> >  . if we are going to add this, it needs documentation
> 
> For sure.  I only posted what I had because the subject came up, not
> because I considered it completely finished.
> 
> >  . why only source file names? the issue is relevant to any file name
> 
> Can you give an example?  My thinking is/was we should keep target paths, host
> paths and source paths separate worlds.  Target path comparisions are handled
> with "set/show target-file-system-kind".  Host paths should be compared with
> other host paths using FILENAME_CMP, which uses the appropriate native
> host compare function.  The source paths setting is for the case of
> e.g., a arm-linux program that was compiled on a Windows machine,
> and is later debugged on an x86-linux host gdb (against a arm-linux
> gdbserver).  You want gdb to consider the target filesystem of unix
> kind, but, you want source path comparisions with paths embedded in
> debug info to be permissive by using dos style path comparisons.

Here's the link to the previous short discussion on this, for the
archives.  Sorry, should have included it earlier:

 <http://sourceware.org/ml/gdb-patches/2010-04/msg00759.html>

-- 
Pedro Alves