You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
Currently, nixseparatedebuginfod fails when started through systemd's socket activation system, because the address was not released for long enough for a normal bind to work.
There are advantages to using socket activation for nixseparatedebuginfod, namely to avoid the indexing process until debug info is actually needed. Would it be possible to provide a configuration option that allows SO_REUSEADDR for use with socket activation?
The text was updated successfully, but these errors were encountered:
I don't think nixseparatedebuginfo is a good fit for socket activation. See previous discussions here Support systemd socket activation #10 I'll add that the debuginfo client from elfutils really likes to cache failures so blocking full minutes until results are available is problematic.
I think that to properly implement socket activation you have to inherit the socket via sd_listen_fds(3). From a cursory look, SO_REUSE_ADDR looks like a workaround.
For these reasons I don't plan work in this direction. If I wanted to dedicate time to reduce the power consumption of nixseparatedebuginfod I'd rather implement watching the nix sqlitedb with inotify, or even remove the need for indexation nixpkgs side.
Currently,
nixseparatedebuginfod
fails when started through systemd's socket activation system, because the address was not released for long enough for a normal bind to work.There are advantages to using socket activation for
nixseparatedebuginfod
, namely to avoid the indexing process until debug info is actually needed. Would it be possible to provide a configuration option that allowsSO_REUSEADDR
for use with socket activation?The text was updated successfully, but these errors were encountered: