About changes in PCB plot output

[Salvador E. Tropea]

Hi All!


One of the things KiBot does is enable repeatable plots, you put the
options in a YAML file and you'll always get plots using these
parameters, regardless of any change in the plot dialog.

To test it my regression tests usually plot PCBs and compare them with
some reference image that was visually validated.

From time to time KiCad changes the plot output. This is quite
annoying, and even dangerous because this includes the gerbers. Note
that the same is for the schematic.

This happens very often and I always wondered if the development team is
aware of this.

Since 7.0.0 release I already changed 3 times the references for KiCad
7, and now I'm facing another change introduced by
d5bd1f5aea042a238345171b99a195da9e0eab9e (which is radical for some
cases, i.e. left justified large text)

Most of the commits that introduced changes claims to be to make things
more compatible with KiCad 6, which is good.

But, if I try to use 6.0.11 images as reference I still have differences
with KiCad 7.0.3, even with d5bd1f5aea042a238345171b99a195da9e0eab9e
applied, so it looks like more of these changes will come.

So here I want to propose something: Can KiCad regression tests do the
same that KiBot does? (make sure the plot outputs doesn't change much)

KiBot tollerates deviations in the color and also some pixels changing.
But here we are talking about hundreads or thounsands of pixels changing
for a 3000x2000 image.

I know this will be annoying because small changes in code will report
errors, but then the team will have to make a decision: Is this change
really needed or is just an unintended side effect?


Please consider it,

Salvador

[Salvador E. Tropea]

I forgot to mention: 7.99 generates files that are already different to
7.0.3 (both from yesterday)

[Mark Roszko]

We are moving towards some unit tests to validate the plot output.
Such a test already exists for schematic plot output as a kicad-cli
test and will be expanded to PCB eventually.

But, even the recent changes such as
d5bd1f5aea042a238345171b99a195da9e0eab9e were entirely intentional and
not an accidental change.
Even the original change that the d5bd1f5 commit is fixing wasn't
entirely accidental either.
But we had a discussion to switch back to the original behavior for
the exact "it's dangerous reason" of opening 6.0 designs.

> --
> You received this message because you are subscribed to the Google Groups "KiCad Developers" group.
> To unsubscribe from this group and stop receiving emails from it, send an email to devlist+u...@kicad.org.
> To view this discussion on the web visit https://groups.google.com/a/kicad.org/d/msgid/devlist/0d745720-dfab-739c-ca4a-76da312d12cd%40inti.gob.ar.



--
Mark

[Salvador Eduardo Tropea]

Thanks for the clarification Mark.

It would be nice if the development team informs about changes that will impact in the plot generation. This could help people to know the changes are intentional.

---

De: dev...@kicad.org <dev...@kicad.org> en nombre de Mark Roszko <mark....@gmail.com>
Enviado: sábado, 20 de mayo de 2023 12:06
Para: dev...@kicad.org <dev...@kicad.org>
Asunto: Re: [kicad] About changes in PCB plot output

 

To view this discussion on the web visit https://groups.google.com/a/kicad.org/d/msgid/devlist/CAJjB1qJbXmf0rNMPp_42XCdAin%2BmRmxG7y0d%2BBWNEFvBcDTrHw%40mail.gmail.com.