Hi. The package builds ok now. I tried all these combinations:
- Machines with 1 CPU and 2 CPUs. - In trixie and also in unstable. - Using the old file backend and the new unshare backend. (That makes 8 combinations, and I tried building the package 25 times in each of them with zero failures in total). My current theory is now that the failures happened only when using the kernel of bookworm. The build logs I have are consistent with that. (The kernel of bookworm is what everybody was using to build packages just before the release of trixie, because autobuilders, either official or the ones used by Lucas or myself, usually run stable). Now, if the package existed in bookworm, still a supported distribution, we could maybe investigate the case where we try to build the package in bookworm from a machine using bookworm (i.e. using the kernel of bookworm). But the package does not even exist in bookworm... so I would say it's fair enough to close the bug at this point (I leave that to you). Thanks.
