bug-guix
[Top][All Lists]
Advanced

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

bug#70926: Having default nss-certs plus nss-certs in operating-system p


From: Maxim Cournoyer
Subject: bug#70926: Having default nss-certs plus nss-certs in operating-system packages causes problems
Date: Wed, 15 May 2024 23:02:27 -0400
User-agent: Gnus/5.13 (Gnus v5.13)

Hello,

Liliana Marie Prikler <liliana.prikler@gmail.com> writes:

> Am Montag, dem 13.05.2024 um 22:38 +0100 schrieb Christopher Baines:
>> I've seen this when updating systems, but it seems like something is
>> wrong with the handling of nss-certs.
>> 
>> I'm on a guix revision with nss-certs by default, and when I add
>> nss-certs to my system packages (to simulate not removing it when
>> upgrading), it breaks certificates (e.g. wget https://guix.gnu.org/
>> doesn't work).
> I can confirm this on three machines (two of my own, one from a
> relative): Having nss-certs in the packages field unexpectedly breaks
> all known certificates.
>
>> My reading of the operating-system-packages code suggests that adding
>> nss-certs shouldn't have any effect, but this doesn't seem to be
>> working.
> It would be really nice to detect the mismatching versions if it's
> based on that.  IIUC we graft nss-certs now, so that we can hot-swap
> stuff like pythons certifi package.  Is this use case broken by any
> chance?

Apparently having multiple nss-certs of the same version is no problem
(they get deduped later).  The original problem would thus only exist
when there are multiple versions of nss-certs listed in packages, as
could happen for installer-generated configs that use
'(specification->package "nss-certs"), which would pick the latest
version and clash with the one in %base-packages.

My code could call delete even in the first case, which would clear
*all* nss-certs because they were the same object.  That's now guarded
against in 35ae95061e1b843e1df069693177519f22f9a16d ("system: Do not
delete all nss-certs packages when they are the same object."), which
I've just pushed.

Closing.

-- 
Thanks,
Maxim





reply via email to

[Prev in Thread] Current Thread [Next in Thread]