[Date Prev][Date Next] [Thread Prev][Thread Next] [Date Index] [Thread Index]

Re: RFC: Make dh_makeshlibs auto-detect udebs in common cases



Niels Thykier <niels@thykier.net> (2019-07-20):
> Thanks for the feedback. :)
> 
> I have tested this and they do.
> 
> Test method: I have verified this against systemd at git master[1] using
> the following three variants/builds:
> 
>  * Vanilla debhelper/12.2.3 with explicit --add-udeb
>  * Patched debhelper/12.2.3 with explicit --add-udeb
>  * Patched debhelper/12.2.3 dropping the "dh_makeshlibs -plibudev1 \
>    --add-udeb=libudev1-udeb [...]" line[2]
> 
> All three generate identical results (i.e. bit-for-bit identical
> debs/udebs)[3].

Thanks, but that's just a single package.

> I will never be able to make code that works all the time given the
> premise that people "blindly" apply changes.  However, it is not my
> goal either. The goal for this change is to simplify udeb handling to
> remove one papercut /in the common case/ for udeb packages in two
> forms:
> 
>  1) Packages that need to add a new udeb will have one less step to
>     worry about when they follow the naming convention for udebs.
> 
>  2) /Some/ existing udeb producing packages will be able to drop their
>     manual overrides (assuming they follow this convention as well).
>     An example of a package that would benefit is bind9, where we can
>     remove about 11 lines[4].
> 
> While most packages seem follow the convention, there are exceptions.
> As an example, I spotted fontconfig[5] which will not benefit from this
> change.  Instead, it (and other exceptions) will have to keep --add-udeb.

While I'm of course fine with the idea of automating what can be
automated, I think I'd be happier to also see some hint somewhere for
maintainers who wish to drop the --add-udeb, e.g. checking the
resulting SHLIBS before/after the change…

The current situation has been fine since I've been dealing with the
installer, filing the occasional bug when a udeb needs to be added
(which would still be needed with your patch applied as adding the udeb
binary would still be needed); and I wouldn't want to have to deal with
“careless” regressions.


IOW: Fine with making maintainers' life easier, but please help them
double check they don't regress (and end up making mine less so).


Cheers,
-- 
Cyril Brulebois (kibi@debian.org)            <https://debamax.com/>
D-I release manager -- Release team member -- Freelance Consultant

Attachment: signature.asc
Description: PGP signature


Reply to: