Hi Simon, On Mon, Aug 06, 2018 at 12:19:39PM +0100, Simon McVittie wrote: > [ discussion on what deprecated means ]
Experience from python2 has shown that deprecation is difficult. You can announce it early and still everyone ignores it. I even try to move packages off deprecated technology where I can (e.g. xmkmf), but converting dbus-glib users is beyond my skills and time budget. Sorry. > not mean people have actually stopped using it, and does not mean that > it can be removed from Debian in a finite time. I think this is key. What I really want is a way forward in finite time. For instance with qt4, there is a deadline (buster). So I don't bother fixing qt4 packages and simply track the individual qt4 removal bugs. Once they move to qt5, I'll notice immediately and try cross building them. For dbus-glib, I waited for your mbf and nothing happened. Presumably, the time is not ripe yet. That's ok. If you see a different way forward, I'm all ears. Long term, the effort is likely better spent at converting the remaining (many) users. > dbus-glib already has a versioned Depends on a better D-Bus implementation > (part of libglib2.0-0). Next time I do an upstream release I'm seriously > tempted to add __attribute__((deprecated("use GDBus instead"))) to > every symbol. Sounds useful to me. It seems that very many downstreams are not fully aware. > If it's sufficiently important to you to justify a trip through NEW, > then I'd prefer to do a MU fixing this and #873617 (there seems little > reason to fix this one but not #873617). Thanks for the patch, I'll use > it as a base. That's not a boolean, but a non-trivial trade-off. When fixing deprecated technology I consider: * How long does it take to get rid of it? * How many users does it have? For dbus-glib the respective answers appear to be "long" and "many". Having many users also means that failures due to dbus-glib shadow other failures. I'm not that much after fixing this bug for its own sake, but for getting a better picture of the remaining failures. So yes, I have come to the conclusion that fixing this is worth the trip through NEW. That's why I sat down, wrote the patch and kindly offered the NMU, not as a threat, but as potential way forward. I gladly honour your preference for doing a MU. I don't even have much urgency here, but I know that a number of people use stable for cross building, so please try to fix it in buster. Thank you for your cooperation in a matter that doesn't look productive to you. Helmut