Re: [PATCH 0/7] Rewrite jevents program in python

[Ian Rogers]

On Wed, May 11, 2022 at 6:56 AM Peter Zijlstra <peterz@xxxxxxxxxxxxx> wrote:
>
> On Wed, May 11, 2022 at 06:50:59AM -0700, Ian Rogers wrote:
> > On Wed, May 11, 2022 at 12:35 AM Peter Zijlstra <peterz@xxxxxxxxxxxxx> wrote:
> > >
> > > On Wed, May 11, 2022 at 12:01:26AM -0700, Ian Rogers wrote:
> > >
> > > > The changes here switch from jevents.c to a rewrite in python called
> > > > jevents.py. This means there is a build time dependency on python, but
> > > > such a dependency already exists for asciidoc (used to generate perf's
> > > > man pages).
> > >
> > > You mean just building perf (not the docs) will now require snake stuff?
> > >
> > > That's very tedious :/ I don't typically have snakes on my machines.
> >
> > Hi Peter,
> >
> > You're right that after these changes python is a build requirement
> > for jevents. We could keep the C code around for the case that python
> > isn't there, but I want to do things like remove the string
> > relocations, sort the events by name so we don't linearly search, etc.
> > which would be a massive chore to keep alive on the C side. An
> > alternative would be to have an empty pmu-events.c file that is used
> > for this case. If you wanted to keep things in C and have jevents like
> > event names, you could use the empty version and link in libpfm4.
>
> I'm not normally linking to libpfm4. All I really care about is that I
> can still build a bare cli perf (not even tui). If all the snake stuff
> is purely optional and it just disables some features, but I do get a
> perf out at the end, then I'm all good.
>

Great! I'll add a conditional "empty" file version for this case.

Thanks,
Ian