Bug#1085124: muon-meson: FTBFS: /bin/sh: 1: /<<PKGBUILDDIR>>/build/muon: Permission denied

[Andrea Pappacoda]

I'd still like to figure out what is really happening here though.

By looking closer at the logs[1], my guess is that the
`subprojects/meson-docs/meson-reference.3` target, which is the 122th
out of 123, runs at the same time as the 123th target, which links
together the final muon executable. So you have:

- target 122 executing build/muon
- target 123 overwriting build/muon with the final version

So target 122 fails to run because build/muon is modified at the same
exact time.

Does this sound reasonable to you?

I'm Cc-ing the muon list to hear their opinion, as this seems to be an
upstream bug, either in the bootstrap script or the build ones.

(for ~lattis/muon subscribers: you can find the relevant Debian bug
report here: <https://bugs.debian.org/1085124>)

Bye!

[1]: https://people.debian.org/~sanvila/build-logs/202410/muon-meson_0.3.0-3_amd64-20241007T065946.200Z

[Santiago Vila]

Post by Andrea Pappacoda

I'd still like to figure out what is really happening here though.

- target 122 executing build/muon
- target 123 overwriting build/muon with the final version
So target 122 fails to run because build/muon is modified at the same exact time.
Does this sound reasonable to you?

That's exactly what it seems, yes.

I think there are two different bugs here.

One of them is a Makefile-type bug (or whatever the equivalent is here,
i.e. I mean the dependencies have not been precisely specified).

The other one is a bug in ninja. Documentation says it will never use
more parallel jobs than available CPUs, but that does not seem to be
true:

$ nproc
1
$ ninja --help 2>&1 | grep "run N jobs"
-j N run N jobs in parallel (0 means infinity) [default=2 on this system]

Note how the default is 2 even if I only have 1 CPU in this machine.

I'll report the ninja bug separately.

In the meantime I suggest to add -j1 to the ninja invocation, as in
the attached patch. The package takes about one minute to build, so
I don't think this will be a problem.

Thanks.