> This particular issue in dpkg is very much known and nothing new (a > short recap: dpkg can lose files if files are moved between packages > *and* symlinked directores, such as / and /usr, at the same time). > > > To mitigate it, bluca added a piuparts check which rejects packages > that move files from / to /usr (for bookworm/sid). This is overly > conservative as strictly speaking, we'd only have to be careful of > packages that move files from / to /usr and between packages within > one release cycle. I.e. it would be perfectly fine to move files for > / to /usr if during a release cycle those files aren't moved between > packages. I suspect this will be a rare case, that said definitely > something to be aware of if we don't get a fixed dpkg. > > Fwiw, once all files are moved to /usr, the dpkg bug is no longer > really relevant. > > So, I think there isn't any new information here in #1020792 which > would warrant a halt. Indeed, there is nothing new reported here, it's a mix of old news - none of the failure modes mentioned can actually happen given the piuparts check that has been in place for a few months (and if someone can prove the opposite please let us know and we'll update it), and baseless, patently false statements - I frankly find it quite upsetting to see claimed that "we have refused to fix any bug" as a self-evident fact, when even a cursory look at the distribution packages/bugs trackers in the past couple of months tells a very different story. Also there were several status updates sent over time, with detailed progress reports, that were linked in the d-d-a mail - pretty hard to miss. In short tons of work has happened, and continue to happen, and seeing it casually dismissed with a shrug is honestly quite disheartening. Finally, extraordinary claims require extraordinary evidence: given it was claimed that "all systems will be rendered unbootable", it should be trivial to show it. Provide the log of an upgrade bullseye -> bookworm that fails to boot. Should be easy enough, given it *allegedly* affects all systems (despite of course nobody ever having seen anything remotely like it, ever, over the course of several years), no? We'll be eagerly waiting for a detailed and evidence-based report. In the meanwhile, I'd humbly suggest close+wontfix. -- Kind regards, Luca Boccassi
Attachment:
signature.asc
Description: This is a digitally signed message part