PSA: Changing bindings generator

[Nils Barth]

Hiyas,

If you get any problems with generating bindings in the build (esp. on non-Linux/ninja), please try syncing to ToT Blink (or at least Blink Revision 150731) and regenerating GYP (gclient runhooks or build/gyp_chromium).

If you still have problems, please open a bug and CC me (nbarth@), and haraken@.

Infra/build experts: are there other tests I should do or warnings I should send?

Background

Per haraken's mail of May 10, we're rewriting the IDL compiler in Python!

Issue 239771: Rewrite the IDL compiler in Python

Indeed, we've just landed the first step:

Issue 241016: Rewrite the IDL dependency resolver preprocess-idls.pl in Python

This does involve some GYP changes, which require regenerating build files sometimes. Unfortunately, these changes also occasionally result in bugs (sorry), which I don't always catch (sorry), due to not testing on every build configuration.

Question: are there other try bots I should be using?

Can additional tests be automatically required for changes in the bindings directory?

...so if you have a build problem, like this bug:

Issue 242154: [win32 chrome build] Unknown option: writefileonlyifchanged

...please open a bug and contact us!

[PhistucK]

So what about the PPAPI and Dart implementations?

Will this effort converge everything?

☆PhistucK

[Anton Muhin]

We plan to migrate Dart generator to Blink's one.

yours,
anton.

[Nils Barth]

Tentatively the Blink IDL parser will be derived a generic Web IDL parser, as will the Pepper and (per Anton) Dart IDL parsers; I'm working on this right now.

Assuming this works, each dialect would only need to implement how much it deviates from the spec, and the front end would be shared as much as possible (back end completely separate of course). Also ideally the dialects won't deviate at all from the standard, or only minimally.

Until this is actually implemented, we won't know for sure, but I have a prototype Blink IDL parser derived from the generic one (in Chromium), so it looks feasible.