Re: request for minor change to canonicalize-lgpl.c

[Eric Blake]

[adding the gnulib mailing list]

On Fri, Aug 06, 2021 at 08:57:05AM -0700, Mike Fulton wrote:
> Hi,
> 
> I am building m4 (latest) on the z/OS operating system and for the most
> part, things are going very well.
> I am not sure of the best place to request this - whether as a 'bug' report
> (which this is not) or as a discussion topic, so I decided on 'discussion'.
> 
> The problem I have hit for z/OS is that canonicalize-lgpl.c is defining
> __stat to stat, i.e.
> #define __stat stat
> 
> Unfortunately, on z/OS, the system header <sys/stat.h> uses __stat as a
> wrapper for the header file to test if the header has been included or not,
> e.g.
> 
> #ifndef __stat
>   #define __stat 1
>   <header file contents>
> #endif
> 
> So - the patch I did was just to use a different name for the macro in
> canonicalize-lgpl.c - I chose __m4stat but any non-colliding name would
> work fine.
> 
> What is the best way for me to proceed - should I open a 'bug' for this?

You already have - I'm forwarding this to the gnulib folks (where the
canonicalize-lgpl.c file comes from), and we will get it fixed to quit
using the identifier __stat for our own uses now that we know system
headers are using it.  Once it is fixed there, I need to do another m4
release anyways (1.4.19 has a bug with unintended locale sensitivity
when it comes to parsing numbers), so eventually when I get the time,
the 1.4.20 release will include the gnulib fix.

-- 
Eric Blake, Principal Software Engineer
Red Hat, Inc.           +1-919-301-3266
Virtualization:  qemu.org | libvirt.org