Re: meson and check-tests

Hi,

On Thu, 26 Sept 2024 at 08:45, Ashutosh Bapat
<ashutosh(dot)bapat(dot)oss(at)gmail(dot)com> wrote:
>
> On Wed, Sep 25, 2024 at 8:24 PM Nazir Bilal Yavuz <byavuz81(at)gmail(dot)com> wrote:
> >
> > Additionally, the patch I shared earlier was only for regress/regress
> > tests. From what I understand from here [1], only regress/regress
> > tests support 'make check-tests', so the patch seems correct. I
> > experimented with how we can implement something similar for other
> > types of tests, including other regression, isolation, and ECPG tests.
> > The attached patch works for all types of tests but only for the Meson
> > builds. For example you can run:
> >
> > $ meson test --suite setup
> > $ TESTS='check check_btree' meson test amcheck/regress
> > $ TESTS='ddl stream' meson test test_decoding/regress
> > $ TESTS='oldest_xmin skip_snapshot_restore' meson test test_decoding/isolation
> > $ TESTS='sql/prepareas compat_informix/dec_test' meson test ecpg/ecpg
> >
> > What do you think about that behaviour? It is different from 'make
> > check-tests' but it looks useful to me.
>
> I think that would be a good enhancement, if a particular regression
> set takes longer e.g. the one in test_decoding takes a few seconds.
> When we worked on PG_TEST_EXTRA, it was advised to keep feature parity
> between meson and make. I guess, a similar advice applies here as well
> and we will have to change make to support these options. But that
> will be more work.
>
> Let's split the patch into two 1. supporting TESTS in meson only for
> regress/regress, 2. extending that support to other suites. The first
> patch will bring meson inline with make as far as running a subset of
> regression tests is concerned and can be committed separately. We will
> seek opinions on the second patch and commit it separately if it takes
> time. It will be good to see the support for running a subset of
> regression in meson ASAP so that developers can save time running
> entire regression suite when not needed. The second one will be an
> additional feature that can wait if it takes more time to add it to
> both meson and make.

I agree with you. I splitted the patch into two like you said.

--
Regards,
Nazir Bilal Yavuz
Microsoft