Mailing List Archive: Re: KDE update puzzle : solved

Re: KDE update puzzle : solved

Sep 19, 2021, 12:03 AM

Post #1 of 5 (375 views)

210918 cal wrote:
>> (kde-frameworks/kglobalaccel-5.82.0-r1:5/5.82::gentoo, installed) USE="-debug -doc -nls -test" ABI_X86="(64)" pulled in by
>> (no parents that aren't satisfied by other packages in this slot)
> See in particular the last message.
> I'm not sure what causes that message in particular,
> but it seems like kglobalaccel is causing the problem
> by pulling in a bunch of 5.82.0 deps causing a slot conflict.

That is the key, but it doesn't work the way you suggest.

I cut the Gordian knot Alexander-fashion :
I simply unmerged the 7 objecting pkgs via 'emerge -C ... ',
ie attica kactivities kconfig kcoreaddons kcrash kglobalaccel kwindowsystem.
after that, I was able to update the whole list of pkgs without a problem.

This has to be a bug in Portage,
at least in the sense of an unintended + obstructive feature.
Portage ought to be able to recognise when a set of pkgs
are creating destructive circular dependencies among themselves
& resolve the conflict by updating all of them together.
This sort of traffic-jam happens regularly & has done so for many years :
one case is Qt, which forces the user to unmerge the whole set,
then merge the updates as a set too.

Perhaps one of our hard-working unpaid devs is listening ... (smile).

--
========================,,============================================
SUPPORT ___________//___, Philip Webb
ELECTRIC /] [] [] [] [] []| Cities Centre, University of Toronto
TRANSIT `-O----------O---' purslowatchassdotutorontodotca

Re: KDE update puzzle : solved [ In reply to ]

antlists at youngman

Sep 19, 2021, 2:53 AM

Post #2 of 5 (375 views)

Permalink

On 19/09/2021 08:03, Philip Webb wrote:
> Portage ought to be able to recognise when a set of pkgs
> are creating destructive circular dependencies among themselves
> & resolve the conflict by updating all of them together.

Until you hit my version of the problem, where if a is not there, b
crashes when you try to build it ...
And of course, installing a crashes because b is not there ...

Cheers,
Wol

Re: KDE update puzzle : solved [ In reply to ]

purslow at ca

Sep 19, 2021, 6:06 AM

Post #3 of 5 (375 views)

Permalink

210919 antlists wrote:
> On 19/09/2021 08:03, Philip Webb wrote:
>> Portage ought to be able to recognise when a set of pkgs
>> are creating destructive circular dependencies among themselves
>> & resolve the conflict by updating all of them together.
> Until you hit my version of the problem,
> where if a is not there, b crashes when you try to build it ...
> And of course, installing a crashes because b is not there ...

So you emerge both together :
I've never run into a version of your case where that wasn't the solution.
And I don't expect anything like that to happen
with a related group of pkgs like Qt or KDE.

--
========================,,============================================
SUPPORT ___________//___, Philip Webb
ELECTRIC /] [] [] [] [] []| Cities Centre, University of Toronto
TRANSIT `-O----------O---' purslowatchassdotutorontodotca

Re: KDE update puzzle : solved [ In reply to ]

cal at mail

Sep 19, 2021, 9:34 AM

Post #4 of 5 (375 views)

Permalink

On 9/19/21 12:03 AM, Philip Webb wrote:
> 210918 cal wrote:
>>> (kde-frameworks/kglobalaccel-5.82.0-r1:5/5.82::gentoo, installed) USE="-debug -doc -nls -test" ABI_X86="(64)" pulled in by
>>> (no parents that aren't satisfied by other packages in this slot)
>> See in particular the last message.
>> I'm not sure what causes that message in particular,
>> but it seems like kglobalaccel is causing the problem
>> by pulling in a bunch of 5.82.0 deps causing a slot conflict.
>
> That is the key, but it doesn't work the way you suggest.
>
> I cut the Gordian knot Alexander-fashion :
> I simply unmerged the 7 objecting pkgs via 'emerge -C ... ',
> ie attica kactivities kconfig kcoreaddons kcrash kglobalaccel kwindowsystem.
> after that, I was able to update the whole list of pkgs without a problem.
>
> This has to be a bug in Portage,
> at least in the sense of an unintended + obstructive feature.
> Portage ought to be able to recognise when a set of pkgs
> are creating destructive circular dependencies among themselves
> & resolve the conflict by updating all of them together.
> This sort of traffic-jam happens regularly & has done so for many years :
> one case is Qt, which forces the user to unmerge the whole set,
> then merge the updates as a set too.
>
> Perhaps one of our hard-working unpaid devs is listening ... (smile).
>

For what it's worth, I have had these same KDE packages listed above
installed for 3 years, and have never experienced the slot conflict you
encountered (in particular, the recent upgrade from 5.82 to 5.85 worked
fine on my systems). So whether or not the end result could be
considered a bug in portage, it seems likely that, like others have
suggested on the thread, there is possibly a strange configuration
somewhere on your system triggering the problem to occur (I have usually
found this to be the case when my emerge updates blow up with strange
conflicts).

I'm not a portage dev, but I suspect if you wanted someone to look into
it, you'd need to file a bug report and clarify the details about your
system profile, world set, USE flags, package keywords, whether the
problem occurs if you emerge @world instead of directly merging certain
packages, etc.

cal

Re: KDE update puzzle : solved [ In reply to ]

antlists at youngman

Sep 19, 2021, 1:53 PM

Post #5 of 5 (375 views)

Permalink

On 19/09/2021 14:06, Philip Webb wrote:
> 210919 antlists wrote:
>> On 19/09/2021 08:03, Philip Webb wrote:
>>> Portage ought to be able to recognise when a set of pkgs
>>> are creating destructive circular dependencies among themselves
>>> & resolve the conflict by updating all of them together.
>> Until you hit my version of the problem,
>> where if a is not there, b crashes when you try to build it ...
>> And of course, installing a crashes because b is not there ...
>
> So you emerge both together :
> I've never run into a version of your case where that wasn't the solution.
> And I don't expect anything like that to happen
> with a related group of pkgs like Qt or KDE.
>
Well, seeing the fact that this particular case is plastered all over
the net with people trying to solve it, then emerging the two together
DOESN'T solve it.

The problem is that - in order to emerge either one - THE OTHER ONE MUST
HAVE BEEN EMERGED FIRST. Whoops ...

(If they're not already on your system, try installing something that
requires truetype. That will call in freetype, which won't emerge
because harfbuzz isn't there. But harfbuzz won't emerge, because
freetype isn't there. As in, they won't emerge even if you try and force
it, because configure will crash because it can't find the other one.)

Cheers,
Wol