Re: Excessive tests in Ruqola / dependencies not built as part of the Craft caches

Le jeudi 16 juillet 2026, 14:23:04 heure d=E2=80=99=C3=A9t=C3=A9 d=E2=80=99=
Europe centrale Ben Cooksley a =C3=A9crit :
> On Fri, Jul 17, 2026 at 12:11=E2=80=AFAM laurent Montel <[email protected]> =
wrote:
>=20
> > Le jeudi 16 juillet 2026, 14:00:23 heure d=E2=80=99=C3=A9t=C3=A9 d=E2=
=80=99Europe centrale laurent
> > Montel a =C3=A9crit :
> > > Le jeudi 16 juillet 2026, 13:35:05 heure d=E2=80=99=C3=A9t=C3=A9 d=E2=
=80=99Europe centrale Ben
> > Cooksley a =C3=A9crit :
> > > > Hi Laurent, all,
> > >
> > > Hi,
> > >
> > > >
> > > > Over the past week or two Sysadmin has received a number of complai=
nts
> > > > regarding a lack of CI builder availability, primarily from those w=
ho
> > work
> > > > on Plasma. Investigation has revealed that Ruqola appears to use a
> > > > disproportionately high amount of CI time compared to the size of t=
he
> > > > project and it's activity.
> > > >
> > > > This appears to be driven by the fact that Ruqola has a large numbe=
r of
> > > > tests that do not take very long to execute (700 or so tests, yet d=
one
> > in 6
> > > > minutes - with many of these being less than 1 second to run). Given
> > that
> > > > every single one of these unit tests needs to have automoc / compil=
er /
> > > > linker run on it i'd suggest that similar tests be merged into the =
same
> > > > compilation unit (you can run the individual tests, but keep similar
> > tests
> > > > together in the same unit test).
> > >
> > > In ruqola each autotest depends against a specific class.
> > > So merging them in one autotest seems not logical.
> > > It's the same idea that merging autotests from kio in one autotest.
> > >
> > > After that I am not against to execute autotests only on linux-qt6-ne=
xt
> > > It's the main target. but I need to know how to told to CI to exclude
> > other platforms.
> > >
> > >
> > > for example for freebsd, I can't debug so I don't need to execute
> > autotest on it.
> > >
> > >
> > > > This would also explain why the build time
> > > > on Windows is significantly longer, as NTFS does not perform well w=
hen
> > > > compared to Linux file systems, even on NVMe storage.
> > > >
> > > > FreeBSD also has 8 tests failing (timing out, with a 1 minute timeo=
ut)
> > and
> > > > Windows has three tests failing with timeout.
> > > >
> > > > If the Ruqola developers could please merge their unit tests into a
> > smaller
> > > > number of unit tests and look into the broken unit tests as a prior=
ity
> > that
> > > > would help your fellow KDE Developers by reducing the wastage of CI
> > time.
> > > >
> > > > On the Craft based builds, Ruqola also uses a stable version of a
> > > > dependency (ktextaddons) that is not in the Craft cache.
> > >
> > > I don't understand why it's not in craft cache as it's used by
> > ruqola/pim*/kaichat.
> > >
> > > I will look at it.
> >
> > When I looked at qt6-dev/BinaryCachePackages.shelf there is ktextaddons
> > same for qt6/BinaryCachePackages.shelf
> >
> > So if I understand it it's ok no ?
> >
>=20
> No it isn't, because you have overridden versions from the default -
> https://invent.kde.org/network/ruqola/-/blob/master/.craft.ini?ref_type=
=3Dheads#L6
> Version overrides should be avoided because they prevent caching of
> dependency builds.
>=20
> The override to ktextaddons is especially bad as Craft has no knowledge of
> that version, so the source tarball is not being validated to ensure it is
> legitimate and has not been tampered with.
>

Ok so problem was not in https://invent.kde.org/sysadmin/craft-ci
I will clean up /.craft.ini

>=20
> >