[LTP] [PATCH] doc: add tests catalog page

Cyril Hrubis chrubis@suse.cz
Mon Feb 3 15:42:01 CET 2025


Hi!
> > I guess that we want group the data somewhat like:
> 
> >  - options (because command line options may be useful there)
> 
> >  - min_kver, needs_kconfigs, min_cpus, needs_drivers, needs_cmds...
> >    (because it may describe why tests are skipped)
> 
> >  - needs_cgroup_ctrls + cgroup knobs (for tests that use cgroups)
> Also .needs_cgroup_ver, .needs_cgroup_nsdelegate

That should be covered by the "cgroup knobs"

> >  - all_filesystems, skip_filesystems, filesystems
> 
> >  - timeout, runtime         (here as well, run for longer)
> 
> >  - needs_root      maybe?
> >  - needs_device    maybe?
> IMHO both can be interesting - you can list tests you loose if you don't use
> root or you don't have kernel with loop device configured.
> 
> Also there are more: .caps, .hugepages, .needs_hugetlbfs, .min_mem_avail,

I guess that .caps is closer to be interanal than not.

> .min_swap_avail, .dev_min_size, .needs_abi_bits, .needs_overlay (at least).

And we also have supported_archs, that is probably interesting.

> Why not to show all the info like it was previously? We never know what the
> reader will be interested in.

The chalenge here is to keep the page nicely formatted and generally
well arranged, having less helps a bit there. So the idea is to skip the
data that are generally mostly interesting to the test runner. The
quesitos is where to draw the line.

> If we are really not interested at all, I would just avoid few LTP true
> internals, e.g.: .needs_checkpoints, .forks_child, .child_needs_reinit.

These are really uninteresting to anone reading that documentation.

-- 
Cyril Hrubis
chrubis@suse.cz


More information about the ltp mailing list