Re: KDE Gear projects with failing CI (master + stable) (11 February 2026)

Stefan Brüns <[email protected]> Mon, 16 Feb 2026 17:19:34 +0100
Newsgroups gmane.comp.kde.devel.general
Message-ID <16559413.hNhHcOVSLA@nibbler>
On Freitag, 13. Februar 2026 22:35:31 Mitteleuropäische Normalzeit Nicolas 
Fella wrote:

> >> 
> >> I don't follow what's going on there.
> >> 
> >> Locally that benchmark finishes in ~1 second here, so no problem
> >> whatsoever.
> >> 
> >> On CI the first two sub-benchmarks seem completely fine, but
> >> voidTestMD5Performance seems to hand/time out. It's not clear to me
> >> how many iterations it does.
> >> 
> >> It looks like QBENCHMARK automatically adjusts the number of
> >> iterations until the results are "good" (whatever that exactly
> >> means). Possibly some CI conditions trigger that to go haywire.

It is often useful to run benchmarks with just a single iteration on CI, to 
make sure it compiles and runs. The actual performance is often irrelevant, 
and varies too much anyway.

Unfortunately, this requires to pass a custom COMMAND with args to CMake's 
add_test, and AFAICS this is not possible to do via ecm_add_test(s).


```
add_executable(mboxbenchmark_bin
    mboxbenchmark.cpp
)
target_link_libraries(mboxbenchmark_bin
    Qt6::Test
    KPim6MBox
)
ecm_mark_as_test(mboxbenchmark_bin)
add_test(NAME mboxbenchmark COMMAND mboxbenchmark_bin "-iterations" "1")

```

Regards, Stefan
signature.asc (application/pgp-signature, 232 B)
-----BEGIN PGP SIGNATURE-----

iHkEABECADkWIQSwWRWIpJbl0W4DemNvf0o9jP6qUwUCaZNDlhsUgAAAAAAEAA5t
YW51MiwyLjUrMS4xMSwyLDIACgkQb39KPYz+qlMqjQCgstltNudX8ygHh2OOLGDp
7UAlfXYAn1owlQMREgRSuvMrITeD/XrNYXuv
=PyJj
-----END PGP SIGNATURE-----