On 7/7/21 2:21 PM, Fabio Valentini wrote:
On Wed, Jul 7, 2021 at 1:38 PM Hans de Goede
> On 7/7/21 1:08 PM, Florian Weimer wrote:
>> * Neal Gompa:
>>> Wait, why don't we have guile 3.0?
>> We have a mandate from Fesco that the core toolchain must depend on
>> Guile. Naturally that makes updates rather difficult.
> So I've gone and checked the Fesco issue where dropping guile
> support from make + gdb was discussed:
> And I must say that I find the argumentation for rejecting the
> change very very weak. I would really expect Fesco to make better
> motivated decisions then this one.
> I'm especially shocked about how Fesco is in essence mandating
> a group of maintainers to spend time maintaining a feature
> where they clearly have indicated they don't want to maintain
> that feature.
> My being shocked here is not so much about the guile issue,
> but about a IMHO much bigger issue underlying this decision:
> Since when does Fesco get to mandate on which features our
> volunteer maintainers get to spend there time ?
> I understand there need to be rules and I can understand
> Fesco denying approval for enabling / adding certain
> features for a wide set of reasons, thus in essence blocking
> volunteers from spending time on something because that
> something is deemed undesirable for Fedora.
> But this is different here Fesco is telling a group of
> maintainers that they must maintain a feature even though
> they have indicated that they find the benefits of that
> feature not worth the amount of time it costs to maintain
> support for that feature. So in essence Fesco is telling
> the maintainers that they MUST spend time maintaining this
> even though they don't want this.
> IMHO this is just outrageous and goes way way beyond the
> purview of Fesco.
> Now if dropping this feature would cause major breakage this
> would a different story, But in the whole discussion about
> this, at least as documented in the Fesco issue, no actual
> users of this feature have been indentified and nothing will
> break by disabling this as far is is known. So since there
> is no known breakage caused by this, I end up circling back
> to this basically telling Fesco that the make/gdb timers
> MUST spend them on maintaining this even though they
> don't want to (and have good reasons for not wanting to).
> Which again, is IHMO pretty outrageous really.
> Sorry Fesco, I know that you all do a lot of (hard) work
> as Fesco members and do your best when making decisions
> like this; and I don't doubt that your intentions where
> well, but you made a big booboo here (IMHO).
> I urge Fesco to reconsider this and I suggest that we
> (Fedora) take another serious look at implementing:
> for Fedora 35.
I feel like I need to chime in here as a FESco member who was part of
As others have already mentioned in this thread, this change was
Rejected by the smallest possible margin, with four +1 votes, two
abstentions, and two -1 votes, while it needed five +1 votes to pass.
As I was one of the two "±0" votes: My reason for not voting +1 was
that impact analysis was only based on anecdata, if at all - either
"popcon data from debian shows that almost nobody has make-guile
installed", or "I can't believe anybody actually uses this", but no
actual analysis was done - which, for a central package like make,
would be important, IMO. The stated reason of "we don't want to
maintain this" was also weak, since the developers still maintain the
guile support upstream, and if this was really in preparation for
dropping guile support from ELN / RHEL 9 via a Change in Fedora 34,
that was not really a good reason to "just do it in Fedora first"
However, you are right - FESCo cannot "force" anybody to maintain
anything, but the change proposal was sufficiently "weak" that it was
- barely - not accepted.
But I think almost any package maintainer in Fedora would actually
*look if any optional feature is actually used* before just silently
dropping it from their package, or at least notify maintainers of
dependent packages that feature X was considered for removal. I don't
see this case any different, just that make is a more prominent
package than most.
This is interesting request. I spent some thinking about it for my
optional package features bind-sdb. Do we have any way to guess what
feature is used at least sometime? Especially hard for make package,
where you cannot expect checking every dependent package by hand.
What would be considered sufficient research about usage of guile? If
package provides it as optional feature among many other features, how
should package owner test one feature is still demanded? Do we have any
best practice? Is asking on users@ and devel@ list enough?
So: I'd like to see actual investigation into whether the guile
support is actually used in any Fedora package, and if it's going to
be removed, it should be removed upstream first.
If it turns out that really actually nobody uses this, why not drop it
upstream, and have the guile support removal come with the next GNU
toolchain Change for Fedora?
Red Hat, http://www.redhat.com/