Hi Santiago,

On  Fr 24 Okt 2025 12:47:20 CEST, Santiago Vila wrote:

* Revisit DEB_BUILD_PROFILES 'noinsttest' and 'nocheck' vs. DEB_BUILD_OPTIONS
     'nocheck': (Closes: #1118264)
- DEB_BUILD_PROFILES='nocheck' -> don't build, don't run and don't build
       the lomiri-content-hub-testability bin:pkg.

Hello. Are you really sure that the above change is correct?

no, I am not 100% sure this is correct.

I believe the spirit of nocheck build profile, in plain English, was
that the "outcome" of building the packages should be the same, and
for that to happen, not only the contents of the packages should be
the same, but also the list of generated packages should also be the
same.

Where does this statement come from? Do we have this written down somewhere? I get the point of it, but lomiri-content-hub source code would need adjustments for this to become possible with the effect of saving CPU time at build time.

Unfortunately, I was unable to find the Debian Policy paragraph
explaining this. Do you have a reference for that?

Nope. I'd be interested in that, too.

I am aware that the wiki has some examples where nocheck is used in
Build-profiles in debian/control, but I'm not really sure that's right
if the spirit of nocheck build profile is just to save some time when
building the packages.

The point is that nocheck does two things in lomiri-content-hub (and in general): it avoids building the unit tests and it avoids running the unit tests. That's the CPU time-saving part it provides.

In lomiri-content-hub, part of the unit test building is producing mock-executables that act as clients against the lomiri-content-hub service during autopkgtests. These mock consumers are packaged as the lomiri-content-hub-testability bin:pkg. The lomiri-content-hub-testability bin:pkg is totally unneeded except for this one purpose. So, for speeding things up at build time (i.e. nocheck), the l-c-h-testability bin:pkg must either be empty (dummy content) or skipped from building. None complies with your above interpretation of what nocheck is supposed to do / to deliver.

To get things more straight one would have to patch upstream's code in a way, that we could build testability executables but omit building the unit tests themselves. But such a patch would be highly Debian specific and I am a bit reluctant to dive into this effort.

So, ideally, we get some more opinions on this, otherwise I will consider this bug as closed. (Sorry, for not being more bureaucratic on this).

light+love,
Mike
--

mike gabriel aka sunweaver (Debian Developer)
mobile: +49 (1520) 1976 148
landline: +49 (4351) 486 14 27

GnuPG Fingerprint: 9BFB AEE8 6C0A A5FF BF22  0782 9AF4 6B30 2577 1B31
mail: [email protected], http://sunweavers.net

Attachment: pgpXW0MPn9khO.pgp
Description: Digitale PGP-Signatur

Reply via email to