Mark Johnston

> @markjdb thanks for this. Nothing jumping out at me, but when I did a code read a couple of days ago I didn't go through that codepath, so it...

The problem is with the `newudata` binding, used to allocate a file watcher object. `_Alignof(bee::filewatch::watch) == 16`, and the constructor relies on this property: (gdb) disas Dump of assembler code...

> @markjdb I don't know why freebsd's clang assumes filewatch::watch is 16-byte aligned. Maybe using [alignas](https://en.cppreference.com/w/cpp/language/alignas) will make clang aware that filewatch::watch is not 16-byte aligned. clang is what decides...

> Well. You can change filewatch's uservalue to 2 if it makes clang work. It turns out that the problem is there no matter how many uservalues there are. Since...

I agree with Igor's view about the container/jail naming. There are lots of interface differences between jails, cgroups, zones, etc.., and it's premature to generalize from a single backend implementation....

@ihoro I believe there is some kind of memory leak in the patch. So far I narrowed it down to something in the FreeBSD sbin tests. If I run them...

> Any detail of the case's environment would be helpful. Let me guess that probably `bricoler` has been used here then probably I will have higher chances to reproduce the...

> @ihoro I believe there is some kind of memory leak in the patch. Hmm, there is something strange going on. It seems to be related to incremental buildworlds that...

> > @ihoro I believe there is some kind of memory leak in the patch. > I will keep digging, but please ignore this for now. Sorry for the noise....

> > I'm pretty sure there is a latent bug in kyua somewhere, related to `-v parallelization`. I've been running the FreeBSD src test suite in a loop with parallel...