Hi, the omission was not intentional, just a result of travel, family and gaveling to remember too many things at once.
I’ll try to do a new upload when the kids fall asleep. I agree that fixing buster is the only priority, I can handle the rest. It should be fairly easy as there’s no new upstream release for 7.0, so I’ll just follow different numbering schemes for external (7.0.33-<1,inf)) vs stretch (7.0.33-0+deb7u<1,inf), so the conflict will be (<< 7.0.33-1~ and it should just work fine. Ondrej -- Ondřej Surý <ond...@sury.org> > On 13 Apr 2019, at 15:00, Ivo De Decker <iv...@debian.org> wrote: > > Hi, > >> On Wed, Apr 03, 2019 at 09:15:28PM +0200, Andreas Beckmann wrote: >>> On 2019-04-03 13:45, Ondřej Surý wrote: >>> Hi Andreas, >>> >>> there will be new upstream release of PHP 7.3 this week, so I’ll handle it >>> as one batch, ok? > > I see you uploaded this new version to unstable without a fix for this bug. > >>> I just wonder if there’s a way how to fix that without breaking >>> co-installability with php7.0-curl from external repositories. >> >> How does such a package look like? What dependencies does it have? >> What's the versioning scheme? > > Do you have a pointer to those external repositories? > > If we can find a nice solution, that would be ok, but if not, I think fixing > this issue in buster is more important than supporting external repositories. > >> We could try a Breaks: libcurl3 in php7.3-common. >> ... trying ... >> Nope, that only make it worse. >> >> Is stretch getting new php 7.0 point releases? In that case we can't >> have a versioned >> Breaks: php7.0-curl (<< external-repo-version-newer-than-stretch) >> that will continue to be valid. (Unless the external repo versions come >> with an epoch) >> >>> What if we fixed this in php7.0-curl package? (I keep separate sources for >>> it.) >> >> I don't think there is much fixable in php7.0-curl, its the >> libcurl{3->4} transition biting us here ... >> >> The big hammer that should work is >> gcc-8-base: Breaks: libcurl3 >> but it's a completely unrelated package ... > > That doesn't seem like a realistic solution. > > Cheers, > > Ivo >