On Mon, Oct 9, 2017 at 1:13 AM, Dominik 'Rathann' Mierzejewski <dominik@xxxxxxxxxxxxxx> wrote: > Dear fellow Fedora developers! > I was puzzled by this after upgrading from F25 to F26 on a headless > machine with a pretty minimal set of packages installed (i.e. almost > no libraries related to Xorg or mesa). The upgrade brought in an > updated abrt package which grew a dependency on > python-gobject/python3-gobject. That, in itself, wouldn't be so bad, > but python-gobject pulls in cairo and over 20 other libraries related > to graphical display. I couldn't find anything in abrt that would depend > on cairo, so I filed: https://bugzilla.redhat.com/show_bug.cgi?id=1499581 . > I don't like to have unnecessary and unused libraries lying around. > > However, this may indicate a deeper issue in pygobject3 or any package > depending on it. It seems that on June 30th, 2015, the non-cairo parts > were split into -base subpackages, but my guess is that not every package > that uses pygobject3 but not cairo switched to the new subpackage. This > is on Fedora 26: It use to be the core non graphics functionality was in pygobject3-base and this has been the case for _YEARS_, the naming might be better (pygobject3 being the base and having the rest in a sub package) but overall this has worked as such for a long time. > $ dnf repoquery --whatrequires python-gobject|wc -l > Last metadata expiration check: 0:15:47 ago on Mon 09 Oct 2017 01:32:28 CEST. > 63 > $ dnf repoquery --whatrequires python3-gobject|wc -l > Last metadata expiration check: 0:15:53 ago on Mon 09 Oct 2017 01:32:28 CEST. > 99 > $ dnf repoquery --whatrequires python-gobject-base|wc -l > Last metadata expiration check: 0:16:01 ago on Mon 09 Oct 2017 01:32:28 CEST. > 34 > $ dnf repoquery --whatrequires python3-gobject-base|wc -l > Last metadata expiration check: 0:16:08 ago on Mon 09 Oct 2017 01:32:28 CEST. > 30 > > I'd argue that it should've been the other way around. Cairo support > should've been split off to a subpackage and any package that required > it should've been updated. Now it's a mess. > > Even pygobject docs say that cairo integration is optional, > (https://pygobject.readthedocs.io/en/latest/guide/cairo_integration.html > and > https://pygobject.readthedocs.io/en/latest/devguide/packaging.html) > and doesn't even list cairo in Runtime dependencies section. > > Frankly, I don't understand why cairo support is forced upon every > pygobject3 user. Unfortunately, the current layout of the package > doesn't allow depending packages to say they depend on pygobject3 > only (unless we can start depending on pythonNdist(foo) already) > or pygobject3 cairo support explicitly in an easy way (the "main" > package which carries _only_ cairo support doesn't indicate this > in any way apart from the file list and its own dependency on cairo. > > So, should the remaining packages that don't need cairo integration > in pygobject3 simply switch to requiring -base subpackage or should > the pygobject3 package be changed instead? I lean towards the latter, > but perhaps there's a very good reason for all of this which I'm > missing. > > Regards, > Dominik > -- > Fedora https://getfedora.org | RPMFusion http://rpmfusion.org > There should be a science of discontent. People need hard times and > oppression to develop psychic muscles. > -- from "Collected Sayings of Muad'Dib" by the Princess Irulan > _______________________________________________ > devel mailing list -- devel@xxxxxxxxxxxxxxxxxxxxxxx > To unsubscribe send an email to devel-leave@xxxxxxxxxxxxxxxxxxxxxxx _______________________________________________ devel mailing list -- devel@xxxxxxxxxxxxxxxxxxxxxxx To unsubscribe send an email to devel-leave@xxxxxxxxxxxxxxxxxxxxxxx
