Re: [LTP] [PATCH v1] Port getxattr03.c to new test API

[Marius Kittler <mkittler@suse.de>]

Am Freitag, 22. September 2023, 14:20:38 CEST schrieb Wei Gao:
> On Fri, Sep 22, 2023 at 11:23:47AM +0200, Marius Kittler wrote:
> > Am Freitag, 22. September 2023, 02:00:32 CEST schrieb Wei Gao:
> > > > #ifdef HAVE_SYS_XATTR_H
> > > 
> > > Normally i saw this kind of ifdef in include/lapi/xattr.h
> > > I suppose we need create include/lapi/xattr.h?
> > 
> > No, we don't need that header here. This macro is actually defined in
> > `config.h` which also makes sense as it is generated at configuration
> > time. Note that also all builds (including musl) are passing with this
> > change, see
> > https://github.com/Martchus/ltp/actions/runs/6259530698.
> 
> Maybe some misunderstanding, i mean put ifdef logic into
> include/lapi/xattr.h instead of define HAVE_SYS_XATTR_H, since i saw ifdef
> logic for judge exist of xxx.h normally handled in include/lapi/xxx.h.

Ah, so I'd add `include/lapi/xattr.h` and to the `#ifdef HAVE_SYS_XATTR_H` 
there. But is that really the way to go in that situation? I've just checked a 
few header files in the `include/lapi` directory and there's no precedence for 
the case when a header is not supported at all and the corresponding test 
should thus be disabled. These headers seem more for abstracting differences 
between different (versions) of C libraries but not for handling the case when 
a test should be skipped completely.

Note that the way I wrote this test was suggested to me in "Re: [LTP] [PATCH 
v1] Port `getxattr01.c` to new test API".



-- 
Mailing list info: https://lists.linux.it/listinfo/ltp