brew cleanup => Bad file descriptor @ rb_file_flock

[tseemann]

I keep getting these errors when I attempt brew cleanup.

That .cache folder contains lots of .brewing files. Most are zero length, but a couple of them have an old version of a Formula.

Any ideas?

% brew cleanup
Warning: Skipping (old) /bio/linuxbrew/Cellar/capnp/0.5.2 due to it being linked
Warning: Skipping gtk+: most recent version 2.24.30 not installed
Warning: Skipping lua: most recent version 5.2.4_3 not installed
Warning: Skipping homebrew/science/snap-aligner: most recent version 0.15 not installed
Warning: Skipping homebrew/science/uproc: most recent version 1.1.2 not installed
Warning: Skipping (old) /bio/linuxbrew/Cellar/ncurses/6.0 due to it being linked
Warning: Skipping homebrew/science/daligner: most recent version 1.0 not installed
Warning: Skipping qt: most recent version 4.8.7_2 not installed
Warning: Skipping (old) /bio/linuxbrew/Cellar/pillow/2.7.0 due to it being linked
Warning: Skipping homebrew/science/seqan: most recent version 2.1.0 not installed
Warning: Skipping jemalloc: most recent version 4.0.0 not installed
Warning: Skipping poppler: most recent version 0.42.0 not installed
Warning: Skipping homebrew/science/kat: most recent version 1.0.7 not installed
Warning: Skipping (old) /bio/linuxbrew/Cellar/vt/0.577 due to it being linked
Warning: Skipping linux-headers: most recent version 3.18.27 not installed
Error: Bad file descriptor @ rb_file_flock - /home/tseemann/.cache/Homebrew/Formula/gmp.brewing

[sjackman]

Hmm. No, I haven't run into that error myself. Perhaps the file system that you're on doesn't support flock? I'm not sure why those brewing files stick around. Should they be deleted automatically?

[tseemann]

My brew is on an NFS mount (shared across nodes).

There does seem to be a bit of noise of flock() problems on Ruby over NFS on modern Centos:
eg. rubygems/rubygems#806

I'll remove all the files and see what happens.

[sjackman]

I've been using Linuxbrew on NFS as well. I haven't had any trouble, but I suspect it depends heavily on the NFS implementation on the server and client.

[sjackman]

Do post here if you find a resolution.