Re: [bitbake-devel][PATCH] cooker: Handle ImportError for websockets

[Richard Purdie <richard.purdie@linuxfoundation.org>]

On Tue, 2025-02-25 at 11:40 +0100, Alexander Kanavin wrote:
> On Tue, 25 Feb 2025 at 11:11, Richard Purdie
> <richard.purdie@linuxfoundation.org> wrote:
> 
> > So whilst I do see your point, we do have to draw a line somewhere and
> > I'm not convinced always using buildtools is the right one as a
> > default, particularly if we want every day developers using the tool
> > too (which we do).
> 
> I'm no longer proposing always using buildtools by default. The point
> of my message was something else: should we direct bitbake-setup users
> to buildtools if they get a fail that can be addressed with them, and
> if so, how? I'm pretty sure a lot of first timers are going to get the
> 'missing python websockets module' problem, and they will find a hint
> about buildtools helpful  (and something that is already installed and
> ready to run even more so). I'd rather think about it now, and not
> when bitbake-setup, standard configs and associated CDN sstate cache
> go 'live'.
> 
> There were several options proposed in my previous message, and they
> were all skipped in response :(

I'd stand by the idea what failure messages should be good in general
and standalone. We have worked on sanity.bbclass in the past to try and
ensure that the messages are understandable and give pointers to users.

What might be missing is documentation in the quick build docs saying
"if you run into host problems, you could use buildtools?"

Perhaps if sanity.bbclass is failing the build, there could be a
message in there mentioning buildtools as possibly helping/

Could bitbake-setup actually run the sanity tests before the build and
wrap the result with a message about buildtools?

Cheers,

Richard