Re: [PATCH 3/3] trace: do not include TCG helper tracepoints in no-TCG builds

[Paolo Bonzini]

On 16/12/20 16:25, Richard Henderson wrote:
> On 12/16/20 9:22 AM, Paolo Bonzini wrote:
> > On 16/12/20 16:13, Richard Henderson wrote:
> > > > @@ -71,7 +71,7 @@ foreach d : [
> > > >                    input: meson.project_source_root() / 'trace-events',
> > > >                    command: [ tracetool, '--group=root',
> > > > '--format=@0@'.format(d[1]), '@INPUT@' ],
> > > >                    capture: true)
> > > > -  specific_ss.add(gen)
> > > > +  specific_ss.add(when: 'CONFIG_TCG', if_true: gen)
> > > >    endforeach
> > >
> > > Sure, unless it makes more sense to conditionalize the entire loop, so that
> > > those targets aren't even created?
> >
> > The targets exist but the files won't be created (the rules will be in
> > build.ninja but they'll be orphan).
>
> Yes, I get that.  So... is there a reason for them to be there, when it seems
> just as easy for them not to be?

Just a dislike for "if"s when conditional add() will do.

Speaking in general, I like the various sections of the meson.build 
files to use either foreach (taking the job of Makefile target pattern 
rules, where "if"s add complication) or if (taking the job of configure 
code, where loops are rare).  It can't always be done, but it helps 
limiting the amount of spaghetti...

Paolo