Re: [LTP] [RFC PATCH 2/2] sched_football: Rewrite into new API

[Petr Vorel <pvorel@suse.cz>]

> Hi!
> > > And the main problem with our build system is not the tooling we choose,
> > > but the complexity imposed by the out-of-tree build implemented in the
> > > complex makefiles. As far as I can tell 99% of the problems would be
> > > solved by ripping out out-of-tree support, which would remove most of
> > > the code we have in there.

> > +1 for removal. I wonder if anybody finds a time to do that.

> That is unfortunatelly a major task...

> > > > But back to the reality, would it be possible to merge this even with broken
> > > > dependency? I'm not sure myself.

> > > I will double check the code, before adding my final reviewed-by.

> > Thank you! It still bothers me that building sched_football does not trigger
> > building a library, but at least it will work for a full build (building whole
> > LTP).

> It's even more subtle, there are more things to consider. The test
> should be rebuild even if any of the headers it includes change, which
> is not detected properly at the moment. For that to work first thing the
> build system has to do is to generate a dependencies with $(CC) -MM and
> then include them in the build system.

Cool, it really detects required headers.
$ gcc -MM  -I../../include tst_needs_cmds04.c
tst_needs_cmds04.o: tst_needs_cmds04.c ../../include/tst_test.h \
 ../../include/tst_common.h ../../include/tst_res_flags.h \
 ../../include/tst_parse.h ../../include/tst_test_macros.h \
 ../../include/tst_checkpoint.h ../../include/tst_checkpoint_fn.h \
 ../../include/tst_device.h ../../include/tst_mkfs.h \
 ../../include/tst_fs.h ../../include/tst_pid.h ../../include/tst_cmd.h \
 ../../include/tst_cpu.h ../../include/tst_process_state.h \
...
 ../../include/tst_cgroup.h

Kind regards,
Petr

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