Re: [PATCH] libxl: avoid golang building without CONFIG_GOLANG=y

[Nick Rosbrook <rosbrookn@xxxxxxxxx>]

On Mon, Aug 24, 2020 at 03:11:41PM +0200, Jan Beulich wrote:
> On 04.08.2020 18:41, Nick Rosbrook wrote:
> > On Tue, Aug 4, 2020 at 12:02 PM Jan Beulich <jbeulich@xxxxxxxx> wrote:
> >>
> >> On 04.08.2020 17:57, Wei Liu wrote:
> >>> On Tue, Aug 04, 2020 at 05:53:49PM +0200, Jan Beulich wrote:
> >>>> On 04.08.2020 17:50, Wei Liu wrote:
> >>>>> On Tue, Aug 04, 2020 at 05:30:40PM +0200, Jan Beulich wrote:
> >>>>>> On 04.08.2020 17:22, Nick Rosbrook wrote:
> >>>>>>> On Tue, Aug 4, 2020 at 10:17 AM Wei Liu <wl@xxxxxxx> wrote:
> >>>>>>>>
> >>>>>>>> On Mon, Aug 03, 2020 at 10:06:32AM +0200, Jan Beulich wrote:
> >>>>>>>>> While this doesn't address the real problem I've run into 
> >>>>>>>>> (attempting to
> >>>>>>>>> update r/o source files), not recursing into tools/golang/xenlight/ 
> >>>>>>>>> is
> >>>>>>>>> enough to fix the build for me for the moment. I don't currently 
> >>>>>>>>> see why
> >>>>>>>>> 60db5da62ac0 ("libxl: Generate golang bindings in libxl Makefile") 
> >>>>>>>>> found
> >>>>>>>>> it necessary to invoke this build step unconditionally.
> >>>>>>>>>
> >>>>>>>>
> >>>>>>>> Perhaps an oversight?
> >>>>>>>
> >>>>>>> This is intentional, and I think the commit message in 60db5da62ac0
> >>>>>>> ("libxl: Generate golang bindings in libxl Makefile") explains the
> >>>>>>> reasoning well. But, to summarize, CONFIG_GOLANG is only used to
> >>>>>>> control the bindings actually being compiled (i.e. with `go build`).
> >>>>>>> However, we always want the code generation script
> >>>>>>> (tools/golang/xenlight/gengotypes.py) to run if e.g.
> >>>>>>> tools/libxl/libxl_types.idl is modified.
> >>>>>>>
> >>>>>>> I hope this helps.
> >>>>>>
> >>>>>> Not really - I'm still not seeing the "why" behind this behavior. I.e.
> >>>>>> why build _anything_ that's not used further in the build, nor getting
> >>>>>> installed? Also if (aiui) you effectively object to the change that
> >>>>>> Wei has given his ack for, would you mind providing an alternative fix
> >>>>>> for the problem at hand?
> >>>>>
> >>>>> Is the solution here to make the target check if IDL definition file is
> >>>>> actually changed before regenerating the bindings?
> >>>>
> >>>> I don't know - Nick? A move-if-changed based approach would likely deal
> >>>> with the r/o source problem at the same time (at least until such time
> >>>> where the directory containing the file(s) is also r/o).
> >>>
> >>> To make sure Nick and I understand your use case correct -- "r/o source
> >>> problem" means you want the tools source to be read-only? But you would
> >>> be fine recursing into tools directory to build all the libraries and
> >>> programs?
> >>
> >> Yes - until we support out-of-tree builds, nothing more can be expected
> >> to work.
> > 
> > Jan - is the problem specifically that a fresh clone,  or `git
> > checkout`, etc. changes file timestamps in a way that triggers make to
> > rebuild those targets? I have not used the move-if-changed approach
> > before, but AFAICT that would be sufficient.
> 
> Since about three weeks have passed and - unless I've missed something -
> the issue is still there, I'd like to clarify who's going to address the
> (how I would call it) regression. I thought I had expressed that if my
> proposed version isn't acceptable, I'd rather see you deal with the issue.
> Did you perhaps imply the opposite?

Hi Jan,

My understanding was that you were going to use move-if-changed to fix
this for now (it seemed everyone agreed this was the quickest short-term fix).

The long-term plan (as George mentioned earlier in this thread) is to have 
a separate repository where the generated Go code is checked-in, rather
than xen.git. While this will eventually fix your problem, we are not
actively working on this yet.

Sorry for any misunderstanding.

-NR