From: Daniel Latypov <firstname.lastname@example.org>
To: "André Almeida" <email@example.com>
Cc: "Shuah Khan" <firstname.lastname@example.org>,
"Brendan Higgins" <email@example.com>,
"David Gow" <firstname.lastname@example.org>,
"open list:KERNEL SELFTEST FRAMEWORK"
"Nícolas F. R. A. Prado" <email@example.com>,
"Marcelo Schmitt" <firstname.lastname@example.org>,
"Leandro Ribeiro" <email@example.com>,
Subject: Re: KUnit Hackathon
Date: Wed, 7 Jul 2021 17:49:14 -0700 [thread overview]
Message-ID: <CAGS_qxpPahxB_b-5FRHpxMj1GoaK_xT3hsQA4TPQCMWfVxrHSg@mail.gmail.com> (raw)
On Wed, Jul 7, 2021 at 5:23 PM André Almeida <firstname.lastname@example.org> wrote:
> Hi Daniel,
> Às 03:48 de 07/07/21, Daniel Latypov escreveu:
> > On Mon, Jul 5, 2021 at 1:41 PM André Almeida <email@example.com> wrote:
> >> Hello,
> >> We belong to two student groups, FLUSP  and LKCAMP , both of which
> >> are focused on sharing kernel and free software development knowledge
> >> and experience with fellow free software developers and newcomers.
> >> As part of our efforts, we'll be organizing a KUnit hackathon in the
> >> next Saturday (July 10), where we intend to help newcomers convert
> >> existing runtime tests (the ones found at lib/) to KUnit and maybe
> >> create new ones. Depending on the number of attendees, a high volume of
> >> patches may be sent throughout the day. We will do our best to review
> >> all patches before they go to the kernel mailing lists hoping to avoid
> >> wasting your time with minor patching issues.
> > That sounds great!
> >> So we wanted to let you know of all this beforehand and give you the
> >> time to send any suggestions or comments on all this. For instance, we
> > Some random, unorganized thoughts:
> > * I (as a kernel newbie) found the ownership of tests under lib/ to be
> > a bit unclear.
> Indeed, but we found that tests under lib/ are the easiest ones to
> convert, given that they deal with more "generic" programming aspects,
> like linked lists and sort, and doesn't requires subsystem specific
Yeah, I think these are the best for a hackathon.
Plus they're a better fit for KUnit right now, since they're the more
standalone bits of code.
Writing tests for more subsystem-specific code can require a lot of
groundwork, e.g. firstname.lastname@example.org is trudging through what needs to
be set in a superblock to write ext4 tests.
Eventually, we'd ideally have some chunks of shared code to setup
subsystem-specific structures for tests, but they don't exist yet.
I didn't articulate it well, but my concern was finding the "right"
people to review the patches.
It feels like an unfortunate risk that we'll have to take, having some
patches potentially sit unreviewed for a while.
> > * Make sure to base on 5.14, there's some new features that could be of use
> > * QEMU support in kunit.py: it's no longer as big of a deal if we
> > can't get tests running under UML. Brendan's made it fairly easy to
> > run tests in an x86 VM.
> > * Skipped tests support: can now mark a test skipped. Some of the
> > lib/ tests could benefit, e.g. test_hash.c
> > * another lib/ test conversion just landed in 5.14, , along with
> > kunit_kcalloc/kunit_malloc_array() helpers
> > * Documentation/dev-tools/kunit/running_tips.rst - has instructions
> > for how to generate code coverage reports using UML/kunit.py
> > * This seems like it could be useful... 
> Thanks for those info! I wasn't aware of those features.
> > * I feel that the hackathon will surface a number of feature requests
> > for KUnit. Some patches might be blocked on these features (like your
> > uuid patch).
> > * Might be good to track these as they come up and note which
> > patches are blocked.
> Good suggestion, I'll try to track this.
Hopefully we can try and help get some of these resolved, but it might
take a while.
> > * Should we make a decision beforehand wrt renaming test files and
> > kconfig options to match the style guide? In the most recent
> > conversion , I left them as-is. The answer depends on the test
> > owner, but we should probably be consistent with what we do in the
> > v1's at least.
> I believe that to left as-is is easier, and we will suggest people to do
> > * I assume test modules used in selftests (test_bitmap, test_printf,
> > etc.) should remain untouched for now ?
> > * KUnit does *not* return a non-zero exit code when you modprobe the
> > test, as the userspace part of those tests assumes.
> > * We could write a wrapper like kselftest/module.sh, but for KUnit
> > test modules, but this extra in-flight patch would become a dependency
> > for maybe multiple other conversions.
> > * I'd like it if we had ^ eventually, but I don't know how others
> > feel about that.
> Yeah, I don't think we would be able to deal with all of this in this
> hackathon, so we won't touch those tests for now.
> > * I had some small patches for converting test_min_heap.c and
> > test_sort.c that I should get around to sending out.
> So this means that we shouldn't convert them, right?
Yeah, you can leave those ones to me.
I'll try and send them out real soon so people can just check the
mailing list to see what's in flight.
I'm not aware of anyone else working on converting the other tests
right now, so everything else should be fair game.
> >  https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/commit/?id=ebd09577be6c15ee2d343cf60e5bb819946a5ee8
> >  I currently see
> > Overall coverage rate:
> > lines......: 16.4% (20034 of 122481 lines)
> > functions..: 18.1% (2032 of 11205 functions)
> > with the following kunitconfig
> > CONFIG_KUNIT=y
> > CONFIG_KUNIT_ALL_TESTS=y
> > CONFIG_DEBUG_KERNEL=y
> > CONFIG_DEBUG_INFO=y
> > CONFIG_GCOV=y
> > This might be a neat way to motivate participants by showing #s going up.
> > Note: I'm more interested in the # of covered functions than anything else.
> > The %s are nice to look at, but not that useful since we're only
> > compiling a small fraction of kernel code.
> Cool! In the end of the hackathon we will compare the numbers to see how
> much we were able to do in the event.
> >> may ask people to add a special tag to their patches so you may batch
> >> review them all at a later time if you wish.
> >> Anyhow, we'd really appreciate having your opinion on this.
> >> Thanks!
> >>  - https://flusp.ime.usp.br/
> >>  - https://lkcamp.dev/
> >> --
> >> You received this message because you are subscribed to the Google Groups "KUnit Development" group.
> >> To unsubscribe from this group and stop receiving emails from it, send an email to email@example.com.
> >> To view this discussion on the web visit https://groups.google.com/d/msgid/kunit-dev/83ff3ed3-e004-29b2-b23c-a2ba4d5a3f7f%40riseup.net.
next prev parent reply other threads:[~2021-07-08 0:49 UTC|newest]
Thread overview: 6+ messages / expand[flat|nested] mbox.gz Atom feed top
2021-07-05 20:41 KUnit Hackathon André Almeida
2021-07-07 6:48 ` Daniel Latypov
2021-07-08 0:23 ` André Almeida
2021-07-08 0:49 ` Daniel Latypov [this message]
2021-07-07 20:23 ` Brendan Higgins
2021-07-08 0:27 ` André Almeida
You may reply publicly to this message via plain-text email
using any one of the following methods:
* Save the following mbox file, import it into your mail client,
and reply-to-all from there: mbox
Avoid top-posting and favor interleaved quoting:
* Reply using the --to, --cc, and --in-reply-to
switches of git-send-email(1):
git send-email \
* If your mail client supports setting the In-Reply-To header
via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line
before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).