Re: Backport of Python 3.6 for Debian Stretch?
Ludovic Gasc wrote: 2018-04-24 22:16 GMT+02:00 Matthew Woodcraft : You didn't say what dependency problem you're seeing. Is it python3.6 depending on python3-distutils? If it is, you might try building the 3.6.5-3 source package from sid rather than the 3.6.5~rc1-1 package from buster, as that dependency has been removed there. That should give you a working Python 3.6, but without distutils which has recently been moved to the python3-stdlib-extensions package. If you need distutils you might try backporting that too. Upgrade the python system from 3.5 to 3.6 might breaks some python system libraries, no ? So long as it's just the binaries from the python3.6 source package, they should install in parallel and cause no trouble (it would be like installing the python3.7 packages in Buster today, which would go in parallel to 3.6). I think rebuilding python3-stdlib-extensions would be OK too if done with enough care; it would need persuading to build packages including libraries for both 3.5 and 3.6. But yes, that's more risky as it means replacing Stretch packages rather than just adding new ones. -M-
Re: Backport of Python 3.6 for Debian Stretch?
Nguyễn Hồng Quân wrote: I'm using Debian 9 on an ARM board (BeagleBone), for our IoT project. We write an application which needs Python 3.6. But we are struggling with packaging Python 3.6 as deb packages for Debian 9. We found the dsc file for Debian buster. But there is difference in dependency hierarchy between Stretch and Buster, so the found script produces *.deb files which are not installable (due to missing dependencies). You didn't say what dependency problem you're seeing. Is it python3.6 depending on python3-distutils? If it is, you might try building the 3.6.5-3 source package from sid rather than the 3.6.5~rc1-1 package from buster, as that dependency has been removed there. That should give you a working Python 3.6, but without distutils which has recently been moved to the python3-stdlib-extensions package. If you need distutils you might try backporting that too. -M-
Profile-guided optimisation / link-time optimisation in Buster
I noticed that profile-guided optimisation and link-time optimisation have both been unconditionally disabled in Buster's Python 3.6 and 3.7 packages, though 3.5 in Stretch had them both enabled (on common platforms). Can anyone tell me why this was done? I see nothing about this in debian/changelog and the indentation in debian/rules looks odd, so I'm wondering if it might even have been an unintentional change. It's these extra lines in debian/rules: with_pgo = with_lto = just before the ifeq ($(with_lto),yes) line. -M-