| Submitter | Jun Wu |
|---|---|
| Date | March 2, 2016, 10:44 a.m. |
| Message ID | <9cd173eff72962ada960.1456915446@x1c> |
| Download | mbox | patch |
| Permalink | /patch/13524/ |
| State | Changes Requested |
| Delegated to: | Yuya Nishihara |
| Headers | show |
Comments
On Wed, 2 Mar 2016 10:44:06 +0000, Jun Wu wrote: > # HG changeset patch > # User Jun Wu <quark@fb.com> > # Date 1456746319 0 > # Mon Feb 29 11:45:19 2016 +0000 > # Node ID 9cd173eff72962ada9601a368efdb8f87e2eed43 > # Parent 855e65accbb4519f7140828efe6800b0b92fd5f4 > chg: drop --cwd / > > We intensionally want chgserver to load repo config to be able to: > - find out what extensions it should load > - detect config changes correctly > Therefore the --cwd flag is dropped. > > diff --git a/contrib/chg/chg.c b/contrib/chg/chg.c > --- a/contrib/chg/chg.c > +++ b/contrib/chg/chg.c > @@ -210,7 +210,6 @@ > const char *baseargv[] = { > hgcmd, > "serve", > - "--cwd", "/", > "--cmdserver", "chgunix", > "--address", opts->sockname, > "--daemon-postexec", "none", I don't like the implicit chdir() introduced by the next patch. Probably we can do "--daemon-postexec chdir:/" instead. It's common for a daemon process to chdir to / once its boot-up sequence is completed.
On 03/03/2016 02:44 PM, Yuya Nishihara wrote: > I don't like the implicit chdir() introduced by the next patch. Probably we > can do "--daemon-postexec chdir:/" instead. It's common for a daemon process > to chdir to / once its boot-up sequence is completed. Sorry for the late response but I am the London oncall this week and was busy handling internal stuffs. I think --daemon-postexec chdir:/ works but is not a good fit because: - For chgserver, chdir / should be always performed, regardless of what the client says. - It needs extra work to make multiple "postexec" actions (say you want to do both chdir and unlink). Turning --daemon-postexec into a multiple flag will make "--daemon-postexec none" look weird. I will skip this one temporarily for now.
On Fri, 4 Mar 2016 11:37:18 +0000, Jun Wu wrote: > On 03/03/2016 02:44 PM, Yuya Nishihara wrote: > > I don't like the implicit chdir() introduced by the next patch. Probably we > > can do "--daemon-postexec chdir:/" instead. It's common for a daemon process > > to chdir to / once its boot-up sequence is completed. > > Sorry for the late response but I am the London oncall this week and was busy > handling internal stuffs. > > I think --daemon-postexec chdir:/ works but is not a good fit because: > - For chgserver, chdir / should be always performed, regardless of what the > client says. No, chdir('/') isn't a requisite. We need it because chg (client) starts a server at arbitrary locations, and the server is daemonized. > - It needs extra work to make multiple "postexec" actions (say you want to do > both chdir and unlink). Turning --daemon-postexec into a multiple flag will > make "--daemon-postexec none" look weird. Yep, "none" looks weird.
Patch
diff --git a/contrib/chg/chg.c b/contrib/chg/chg.c --- a/contrib/chg/chg.c +++ b/contrib/chg/chg.c @@ -210,7 +210,6 @@ const char *baseargv[] = { hgcmd, "serve", - "--cwd", "/", "--cmdserver", "chgunix", "--address", opts->sockname, "--daemon-postexec", "none",