On Wed, Jul 29, 2020 at 12:00:46PM +0200, Peter Krempa wrote:
On Wed, Jul 29, 2020 at 11:47:31 +0200, Pavel Hrdina wrote:
> On Wed, Jul 29, 2020 at 10:29:46AM +0100, Daniel P. Berrangé wrote:
> > On Wed, Jul 29, 2020 at 09:11:11AM +0200, Pavel Hrdina wrote:
> > > So I was trying to figure out what to do with our syntax-check and this
> > > could be one solution:
> > >
> > >
> > > rc = run_command(
> > > 'sed', '-n',
> > > 's/^\\(sc_[a-zA-Z0-9_-]*\\):.*/\\1/p',
> > > meson.current_source_dir() / 'syntax-check.mk',
> > > check: true,
> > > )
> > >
> > > sc_tests = rc.stdout().strip().split()
> > >
> > >
> > > This is how syntax-check.mk gets the list of targets to run for
> > > syntax-check target. We can use the same list to define tests like this:
> > >
> > >
> > > foreach target : sc_tests
> > > rc = run_command(
> > > python3_prog, '-c',
> > > 'print("@0(a)".replace("sc_",
""))'.format(target),
> > > check: true,
> > > env: runutf8,
> > > )
> > > name = rc.stdout().strip()
> > >
> > > test(
> > > name,
> > > make_prog,
> > > args: [ '-C', meson.current_build_dir(), target ],
> > > depends: [
> > > potfiles_dep,
> > > ],
> > > suite: 'syntax-check',
> > > )
> > > endforeach
> >
> > I like this idea as it eliminates a little bit more of the "make"
> > usage. BTW, can we just run them more directly instead of via
> > "python_prog" ? The tests don't use python, so avoiding creating
> > a python intepretor for each syntax check rule probably wins for
> > performance a litle
>
> The run_command() using python3_prog is executed during `meson setup`
> phase and the only purpose of that is to rename `sc_test_name` to
> `test_name`. It will not affect the performance of running meson test
> as that one will execute only `make_prog -C builddir sc_test_name`.
>
> I'm OK with dropping the run_command() part completely, which would make
> the output of `meson test` look like this:
>
> 147/154 libvirt:syntax-check / sc_prohibit_test_double_equal OK
I think this is okay. It even helps you finding the checking rule.
>
> instead of
>
> 147/154 libvirt:syntax-check / prohibit_test_double_equal OK
>
> I just realized a huge drawback of this approach. In order to run
> `meson test` or `ninja test` it will first compile everything. It can be
> disabled by running `meson test --no-rebuild` which will ignore explicit
> dependencies as well.
>
> To workaround it for our CI codestyle job we would have to run these
> commands:
>
> meson build
> ninja -C build libvirt-pot-dep
> meson test -C build --suite syntax-check --no-rebuild
IMO we can live with this. CI can be fixed. Downstreams which
specifically care about backporting non-conformat patches will probably
disable the syntax-check suite anyways. For developers, you usually
compile stuff anyways before running tests.
Perfect, I'm OK with all of the above so I'll go with it.
Thanks
Pavel