On Wed, Jul 19, 2023 at 6:11 AM Peter Robinson <pbrobinson(a)gmail.com> wrote:
On Wed, Jul 19, 2023 at 10:21 AM Jaroslav Mracek <jmracek(a)redhat.com> wrote:
>
> > On Wed, Jul 19, 2023 at 6:23 AM Jaroslav Mracek
<jmracek(a)redhat.com> wrote:
> >
> > Does that mean the issues with dnf [2] we able to be solved all the
> > time but just weren't investigated?
>
> The issue was investigated also with DNF, but the issue was well hidden, because the
code uses hard coded set for downloaded elements. For most investigation we used the
biggest repository (Fedora) that showed a high memory usage and we tried to mitigate what
can we do to improve the situation. The real issue was with update repository that
surprisingly uses slightly more RAM then fedora repository.
>
> With DNF5 we reinvest it as a completely different issue. DNF5 has a better option
for investigation that allow us to discover the real source of the issue. We knew that
DNF5 fixed RAM usage for `fedora` repository therefore we continued to search in other
directions. Basically we were surprised why we got the report with DNF5 because we know
that RAM usage was improved with DNF5 and default setting. It means that there where two
issues that overlaps with symptoms but has a different reproducers. Solving the first one
(too big metadata to process) uncover the second issue with processing updateinfo
metadata.
>
> The status of the issue - We have to wait until our patch is reviewed and merged in
libsolv and we have to wait until libsolv creates the upstream release, because downstream
of libsolv in Fedora is not under DNF team control and the main admin doesn't like any
downstream patches.
Looking at upstream releases it seems they don't release often, in the
last 18 months there's been 4 releases anywhere between a month and 9
months apart.
I don't see how it's feasible to sit around and tell users "I'm sorry,
you have to wait until upstream bothers to release before you can have
a fix to enable you to update your system" when there is a fix
available. Can you please explain that to me? It is entirely
reasonable to pull in a fix that is headed upstream to fix a key
problem in a key distro component so that it doesn't remain broken for
MONTHS!
I agree. But the reason releases don't get made is that libsolv
doesn't have a set schedule, and I can just ask upstream to make a
release and they probably will.
If the fix is already merged upstream, it's reasonable enough to
backport. What we want to avoid is non-upstream patches, because this
component is critical enough that we don't want that burden.
--
真実はいつも一つ!/ Always, there's only one truth!