On 04/14/2017 08:54 PM, Todd Rme wrote:
On Tue, Apr 11, 2017 at 8:21 PM, Simon Lees <sflees@suse.de> wrote:
I don't see a reason why the singlespec approach can not be used just because the package is developed elsewhere also I can not answer that question.
If the maintainers of said devel project won't/can't support the singlespec approach for the time being, I'd say it's a problem.
So, in that case, would python3- packages be acceptable in d:l:p?
My simple guess would be 'no' because d:l:p has (or should have) the simple requirement to only accept new packages when they follow python singlespec. If there is no maintainer to support the single spec recipe, why should it be in d:l:p?
This depends greatly on the package, say it is a graphical PDF viewer for example, or any other GUI package if it is not designed to have any modules to share with other applications there is simply no need to convert it to single spec, in 5 years when the maintainer decides its right they can just swap the 3's to 4's at there choosing and be done. All the python-efl apps in X11:Enlightenment:Factory are like this and will probably stay that way because there is no need.
Agreed, in fact the singlespec system doesn't even support this use-case. Packages must by named "python-foo", and openSUSE policy forbids packages like this from following that naming scheme.
Well the actual python-efl package is called python3-efl :-) but the applications are all just there own names ie the PDF Viewer called Lekha is in a package called lekha.
Where as libraries that only currently build and support python3 I guess you can make a case that its ideal for them to swap to singlespec now in case we start supporting another python3 impl and that way porting to python4 would be slightly easier but I'm not sure that its a pressing enough issue that we should drop package X from openSUSE:Factory because its in d:l:p3 which will be removed at some point, and know one has put in the effort to convert to single spec, it would be nice if it was converted but for all intents and purposes it still will work fine atm.
As I said previously, I think the last step in the singlespec conversion process would be for these to be moved to d:l:p and have d:l:p3 deleted entirely. These last packages can then be eventually converted at everyone's' convenience but it it wouldn't be a pressing issue. Although of course it will become a pressing issue once we get pypy3 working.
We probably instead should have a bigger focus on depreciating packages in d:l:p that only build with python2 and don't look like they will be migrated to 3.
Why? If the package works, why should we do extra work to remove it? I can understand that we shouldn't do anything excessive to fix such packages once they break, (as long as there aren't potential security issue), but I also wouldn't actively remove them as long as they work and won't cause confusion (particularly with packages that have up-to-date forks).
This was more a statement about longer term planning then anything else, with the fact that python2 will go away and for planning sake we should be thinking about a d:l:p with no python2 and giving it far more weight then thinking about a d:l:p that is still mostly python2 or singlespec. -- Simon Lees (Simotek) http://simotek.net Emergency Update Team keybase.io/simotek SUSE Linux Adelaide Australia, UTC+10:30 GPG Fingerprint: 5B87 DB9D 88DC F606 E489 CEC5 0922 C246 02F0 014B