On Tue, Nov 19, 2019 at 8:22 PM Adam Jackson <ajax@redhat.com> wrote:
On Fri, 2019-11-15 at 23:38 +0100, Kevin Kofler wrote:
> Adam Samalik wrote:
> > 1/ A history chart for base images [2] is now being generated — includes
> > data since 25 September. It's a bit rough initial implementation, but it's
> > there!
>
> Almost 2 months of work to save… 0.5%! That does not look like a huge
> success to me. You simply cannot win against the creeping bloat. :-(

Ahem.

In the spirit of positivity and collaboration, I spent a few minutes
looking at the results given to try to find some easy wins. Here's what
I found:

python3-libs ships multiple copies of its pyc files, corresponding to
different optimization levels. I don't know what a good packaging
solution to that would look like, but if we only shipped the -O2 kind
(which seems appropriate for minimization, as they're smallest) we
could drop about 13M out of 32M, which seems pretty great.

About 2/3 of glib2 is translations. If it was langpacked like glibc you
could drop another 8M. Likewise about 9M out of 10M for coreutils-
common, 4.5 out of 7.5 for bash, 4 out of 10 for gnupg2.

You're not using coreutils-single, which would save you another 6M or
so.

It's hard to understand why dejavu-sans-fonts is in there, since there
are zero font libraries installed in the container base. Probably
that's brought in by the langpacks somewhere along the line, but it
seems like fonts and translations should be logically separate here (no
point installing fonts if you're not going to be printing or making
images, right?). That's another 5.5M.

That's about 44M worth of potential savings out of a 204M base image, a
bit over 20%. I'll happily file proper bug reports for these somewhere
if we want, but it took me like 30 minutes to look into this. If you're
not even willing to put in _that_ little effort, then forgive me for
not taking your assertion that "you simply cannot win" very seriously.

Hey Adam,

Thanks for looking into that. If you could file those bugs that would be fantastic.

I'm sure we can win if there are enough eyes looking for improvements :-) Especially if we have tooling / services that will help us to _keep_ things over time.

Cheers,
Adam

 

- ajax
_______________________________________________
devel mailing list -- devel@lists.fedoraproject.org
To unsubscribe send an email to devel-leave@lists.fedoraproject.org
Fedora Code of Conduct: https://docs.fedoraproject.org/en-US/project/code-of-conduct/
List Guidelines: https://fedoraproject.org/wiki/Mailing_list_guidelines
List Archives: https://lists.fedoraproject.org/archives/list/devel@lists.fedoraproject.org


--

Adam Šamalík
---------------------------
Senior Software Engineer
Red Hat