Re: [PATCH v2 3/3] libtracefs: Add unit tests for user_events

[Beau Belgrave <beaub@linux.microsoft.com>]

On Wed, Feb 23, 2022 at 10:17:15AM -0500, Steven Rostedt wrote:
> On Tue, 22 Feb 2022 15:23:16 -0800
> Beau Belgrave <beaub@linux.microsoft.com> wrote:
> 
> > Adds unit tests for user_events when available. Ensures APIs are working
> > correctly and appropriate errors are being returned.
> > 
> > Signed-off-by: Beau Belgrave <beaub@linux.microsoft.com>
> > ---
> >  utest/tracefs-utest.c | 233 ++++++++++++++++++++++++++++++++++++++++++
> >  1 file changed, 233 insertions(+)
> > 
> 
> OK, so I couldn't get the selftest working because I didn't have
> user_events.h. I then worked to get that, but the selftest still failed
> with:
> 

The suspense is killing me, what did it fail with? :) (Guessing you saw
the fprintf message when I simulate a failure).

> After thinking about this a bit, I've decided that I'll release 1.3 without
> this patch series.
> 

No worries.

> The reason being, I want 1.3 to get into distros ASAP. As adding this patch
> series will put a dependency on user_events.h (yes it builds without it,
> but distros want everything that it can build applied), then that puts a
> dependency on 1.3 to having user_events.h available, which may be a while
> as it has to get into 5.18, and then slowly move to the distro kernels.
> 

Understood.

> If distros did build it without user_events.h and then later on with
> user_evnets.h then we have two versions of 1.3 where one supports
> user_events and one does not. And you can not use versioning to determine
> if your application will link to the library or not.
> 
> With all this in mind, I've decided to hold off this to libtracefs 1.4, and
> when user_events is solidly in the kernel.
> 
> But I'm very excited to have this work in both the kernel and libtracefs. I
> just want it properly done though.
> 

Yeah, do you want me to send out a v3 with some of the fixes? Or have
you put this into a branch somewhere already (I saw your other patch on
top of this one). I fixed some style nits that were bugging me, as they
didn't align to the kernel style.

Thanks,
-Beau