Did we never get SO_REUSEPORT? illumos.org/issues/12455

Anyone get this to work? grpc/grpc #17748. third_party/cares/cares/src/lib/ares_setup.h:32:10: fatal error: ares_config.h: No such file or directory

Smithx10 (LIBERA-IRC): Nope, came across this when trying to run caddy in LX. It's patched out for native github.com/NetBSD/pkgsrc/blob/trunk…dy/patches/patch-listen__illumos.go

Smithx10 (LIBERA-IRC): ^ SO_REUSEPORT

I tried to reach out to Hokuto Araragi about their patch but did not get an answer.

Was trying to run gitlab in a native zone and ran into that ruby issue ./shucks

Smithx10: we provide both grpc and libcares in pkgsrc, does it not work against those?

happy to help through getting it work, native is always preferable

ahhh, i didn't see it in pkgsrc

gem install grpc is what gitlab is puking on

jperkin: gist.github.com/Smithx10/f5aa2688e37a5fea87099eaa472165b9

looks like in this ticket they were open to getting it fixed but that user bailed grpc/grpc #17748

jperkin: so I think ruby is just not looking at the system libs at all and only at the third_party

Anyone ever see this before? gist.github.com/Smithx10/ca090d2e9947f51f5578ffd68753d150 can't zlogin into instance, but it has space

malloc error means it's out of ram.

Resize it and try again.

Also, it's a really good idea to reboot it. If it's been running out of memory I wouldn't trust the state of the applications running in it for business transactions.

Smithx10: ok so looks like we need to add a ruby-grpc package - also there's an existing gitlab package in wip but it may need some work

Anyway to have json exit 1 if its piped input is empty?

An empty string is valid JSON

k

jperkin: would it be easier to have the grpc folkes in that issue just fix the illumos gem? Did you by chance open up the image build repo?

I have a service that has a tendency to create files and then instead of letting the setgid bits on the directory DTRT, it chowns them incorrectly. I've pursued a variety of options for dealing with this, and have come to the conclusion that the only really viable one is to eat kernel audit records to watch for events of interest. Catch is that though I'm watching the fc and fm classes, I ...

... never actually seem to see files getting created or chowned/chmodded in the relevant tree. I do see a bunch of other things that make sense in those two classes. Can anyone help me figure out how I'm screwing this up?

Smithx10: they are somewhat orthogonal, it's nice for stuff to build from source, but my priority is always to provide easy-to-install binary packages so you don't have to

Smithx10: will discuss the repo thing this week, bank holiday here today

ahhh nice, whatever works.... I'm not sure what else I'll hit while trying to compile gitlab

Since gitaly came out, seems like the only HA git / build server that folks are into now