Horton

Thread

Commits

  1. oauth: Correct test dependency on oauth_hook_client

  1. [OAuth] Missing dependency on oauth_validator test

    Jonathan Gonzalez V. <jonathan.abdiel@gmail.com> — 2026-01-17T17:20:55Z

    Hi,

While working in another patch[1] I found that the test for the
oauth_validator was failing when running the test using:

meson test -C build --setup --suite oauth_validator

This was due to the binary `oauth_hook_client` being missing because I
build the code without enabling the OAuth test using `PG_EXTRA_TEST`.

I'm attaching a small patch that add the binary dependency in the
meson.build file, thus, it will be build before running the test.


Regards,

[1] https://commitfest.postgresql.org/patch/6183/

-- 
Jonathan Gonzalez V. <jonathan.abdiel@gmail.com>
EnterpriseDB

  2. Re: [OAuth] Missing dependency on oauth_validator test

    Jacob Champion <jacob.champion@enterprisedb.com> — 2026-01-27T20:17:21Z

    On Sat, Jan 17, 2026 at 9:31 AM Jonathan Gonzalez V.
<jonathan.abdiel@gmail.com> wrote:
> I'm attaching a small patch that add the binary dependency in the
> meson.build file, thus, it will be build before running the test.

Pushed and backpatched, thanks! I missed the conversation in [1],
where this was fixed for the other binaries.

Andres, do you know of any way we could surface these missing deps
with only the "standard" meson/ninja invocations? The only idea I had
was to not build the test executables by default, but I think that
trades a small problem (we occasionally hit a build-time dependency
bug) for a big one (devs fail to notice that a test executable doesn't
compile).

--Jacob

[1] https://postgr.es/m/qh4c5tvkgjef7jikjig56rclbcdrrotngnwpycukd2n3k25zi2%4044hxxvtwmgum

  3. Re: [OAuth] Missing dependency on oauth_validator test

    Andres Freund <andres@anarazel.de> — 2026-01-27T22:39:57Z

    Hi,

On 2026-01-27 12:17:21 -0800, Jacob Champion wrote:
> Andres, do you know of any way we could surface these missing deps
> with only the "standard" meson/ninja invocations?

I don't see a realistic magic solution here, unfortunately.


> The only idea I had was to not build the test executables by default, but I
> think that trades a small problem (we occasionally hit a build-time
> dependency bug) for a big one (devs fail to notice that a test executable
> doesn't compile).

I was thinking that we ought to do that, just to make the compile-hack loop a
bit shorter.  If we then run the tests in some CI tasks after cleaning the
build tree, it should find most of the omissions.

Greetings,

Andres Freund