Re: [PATCH i-g-t 0/7] igt: Help out coccinelle

["Ville Syrjälä" <ville.syrjala@linux.intel.com>]

On Mon, Nov 24, 2025 at 07:12:12PM +0100, Peter Senna Tschudin wrote:
> Hi Ville,
> 
> I need to see the value proposition before supporting ~2000 lines of churn.
> 
> - Can you share the specific Coccinelle scripts that require these changes?

It doesn't matter what you want to change or how. Any time cocci sees
eg. igt_fixture or igt_main it barfs and gives up. So currently you
can't do any kind change with coccinelle to most of tests/.

> - What transformations are you planning that can't work with the current
> syntax?

We need to fix the massive pipe vs. crtc mess. And in general
it would be nice to be able to use cocci over the igt codebase.
Currently that is impossible.

> - Have you attempted these transformations and confirmed current syntax
> blocks them?

Of course I tried, How else do you think I could have
even come up with this stuff?

> 
> More importantly: Coccinelle provides mechanisms to handle non-standard
> macros (iterator declarations,

Iterators is exactly what I used. Did you read the
commit messages?

> custom isomorphisms, virtual rules,
> etc.). Have you exhausted all options for teaching Coccinelle about our
> existing macro syntax before concluding we need to modify the codebase
> itself?

"all options" would probably involev rewriting coccinelle. So no, I
haven't exhausted all options. I have spent quite a bit of time with
coccinelle and I'm not aware of any way to teach it about this kind of
non-standard syntax.

If you can pull some kind of rabbit out of your hat please do so,
and stick it into scripts/ so that other people can also use cocci
without spending a month reverse engineering the syntax.

> 
> It seems backwards to change ~2000 code locations to accommodate a tool

It doesn't really matter how many places need changing. It's all
done completely mechanically with sed, so entirely trivial.

> when the tool is designed to be configurable for exactly these situations.

It's designed to consume somewhat standard C. igt is not that.

> 
> Without concrete examples of problems being solved and evidence that
> Coccinelle-side solutions were explored, this looks like speculative
> refactoring with unclear ROI.

If you need to convince yourself then just try to do some kind of
trivial change over all of tests/ and see how far you can get
without this series and its scripts/iterators.cocci.

-- 
Ville Syrjälä
Intel