| commit | 63ab07d9e08c868121a2721442bb07018d812d84 | |
| author | Jez Ng <jezng@fb.com> | |
| Wed, 3 Feb 2016 21:42:25 +0000 (3 13:42 -0800) | ||
| committer | hhvm-bot <hhvm-bot@fb.com> | |
| Wed, 3 Feb 2016 22:00:46 +0000 (3 14:00 -0800) | ||
| tree | b4d2d30e78a3ca87bf2bcae3842bbec6cbfc56ae | treesnapshot (tar.gz zip) |
| parent | 3f8842fab4049041a36c8b808a2c38621a9c6bea | commitdiff |
Client wakeup should be handled by monitor
Summary:
While working on the exit-on-NFS diff, I noticed that hh_client would hang
forever if the typechecker process exited. It was waiting on a wakeup message
from the typechecker that would never arrive.
I realized that there was no need for the typechecker process to send the
wakeup message -- it's only necessary before the server socket is ready. Now
that the monitor process handles the socket connection, we can have it wake up
the client as well.
This *does* break `hh_client start --wait`, but I think it's a pretty redundant
flag: if the user wants to wait for hh_client to start up, they can simply call
`hh_client check`. I've replaced it with a deprecation message.
Reviewed By: alexchow
Differential Revision: D2884554
fb-gh-sync-id: 7fcb4e54edba131394ca527e6b96472bcbb97a51
Summary:
While working on the exit-on-NFS diff, I noticed that hh_client would hang
forever if the typechecker process exited. It was waiting on a wakeup message
from the typechecker that would never arrive.
I realized that there was no need for the typechecker process to send the
wakeup message -- it's only necessary before the server socket is ready. Now
that the monitor process handles the socket connection, we can have it wake up
the client as well.
This *does* break `hh_client start --wait`, but I think it's a pretty redundant
flag: if the user wants to wait for hh_client to start up, they can simply call
`hh_client check`. I've replaced it with a deprecation message.
Reviewed By: alexchow
Differential Revision: D2884554
fb-gh-sync-id: 7fcb4e54edba131394ca527e6b96472bcbb97a51