Re: [patch] compile: rm -rf -> ftw()+rmdir()+unlink() [Re: [patch] compile: Fix MinGW build]

[Pedro Alves <palves@redhat.com>]

On 12/18/2014 09:23 PM, Jan Kratochvil wrote:
> On Thu, 18 Dec 2014 22:11:16 +0100, Eli Zaretskii wrote:
>> If this is a problem, how about writing down the requirements?
> 
> I was more general above.  In this specific case it would mean describing APIs
> of all the supported OSes.  I do not think it is possible.

- gnulib seems to be good at describe portability issues.
- We use gnulib as low level host portability layer

Always check gnulib.

> This can be implemented by a buildbot for all supported OSes.

Yes, we're getting there.  We need a working master buildbot setup first.
Ideally, that'd be setup in sourceware.org itself, and then we'd
use machines in the gcc compile farm for the build slaves.  But,
as you know, we're actively working on this.

> A prerequisite for that is to settle down a list of really supported OSes.

Not necessarily.  I see it the other way around.   Once buildbot is up
and running for a while, we can require that either people setup
buildbot slaves for their favorite host OSs, or the OSs will be considered
unsupported.

Meanwhile, we should start with a list of OSes gdb supposedly
builds for, per gdb/configure.host.  I had built this table
a while ago:

 https://sourceware.org/ml/gdb/2014-10/msg00026.html

It's obviously quite bare, missing OS version info, for
example.  I had extended it since with GDBserver info,
but left it to neglect on my hard drive since.  But it's a start.
I've now put it up on the wiki, here:

  https://sourceware.org/gdb/wiki/Systems

(linked from the home page).

Maybe once this has more info, it should be migrated to
MAINTAINERS or README in the sources, but while in "collect info"
phase at least, wiki is easier.  Everyone, please do feel free
to extend|massage|whatever this page.

Thanks,
Pedro Alves