[PATCH 1 of 4 V3] chgserver: implement validate command

Yuya Nishihara yuya at tcha.org
Sun Mar 6 09:00:10 EST 2016

On Sun, 6 Mar 2016 12:38:45 +0000, Jun Wu wrote:
> On 03/06/2016 10:26 AM, Yuya Nishihara wrote:
> > I noticed the inconsistency between self.ui and self.hashstate.confighash,
> > but that's okay so long as the server process is forked per request. If the
> > server process is permanent, and if the client ignores the "redirect"
> > instruction and send "runcommand", new extension would be loaded.
> Right. The behavior relies on "fork". I don't think we can switch to
> threads easily because SIGINT will be much harder to handle correctly.

Threading won't be the way to go, but pre-forking will be doable if we can
get rid of globals.

Perhaps the problem can be avoided by discarding new ui if hash differs.

More information about the Mercurial-devel mailing list