procutil: always waiting on child processes to prevent zombies with 'hg serve'
When runbgcommand is invoked by an extension with ensurestart=False, we never
called waitpid - which is fine in most cases, except if that's happening on a
command server (e.g. chg), in which case the child defunct process will just
sit there for as long as the server is running.
The actual semantics of SIGCHLD signal handling is a lot more complex than
it seems, and the POSIX standard *seems* to read that it's ignored by default
and everything would just work without the waitpid if we're not listening for
it, but the truth is that it's only ignored if we *explicitly* set it to
SIG_IGN. We further cannot set it to SIG_IGN or to a catch-all handler across
all of 'hg serve', because Python's suprocess.Popen relies on that signal,
and a few specific parts of hg also set custom handlers, so instead we wait
for specific PIDs in dedicated threads.
I did a poor-man's benchmark of the thread creation and it seems to take
about 1ms, which is way better than the 20+ms from ensurestart=True.
Differential Revision: https://phab.mercurial-scm.org/D8497
== New Features ==
* Windows will process hgrc files in %PROGRAMDATA%\Mercurial\hgrc.d.
== New Experimental Features ==
== Bug Fixes ==
* The `indent()` template function was documented to not indent empty lines,
but it still indented the first line even if it was empty. It no longer does
that.
== Backwards Compatibility Changes ==
== Internal API Changes ==
* Matcher instances no longer have a `explicitdir` property. Consider
rewriting your code to use `repo.wvfs.isdir()` and/or
`ctx.hasdir()` instead. Also, the `traversedir` property is now
also called when only `explicitdir` used to be called. That may
mean that you can simply remove the use of `explicitdir` if you
were already using `traversedir`.
* The `revlog.nodemap` object have been merged into the `revlog.index` object.
* `n in revlog.nodemap` becomes `revlog.index.has_node(n)`,
* `revlog.nodemap[n]` becomes `revlog.index.rev(n)`,
* `revlog.nodemap.get(n)` becomes `revlog.index.get_rev(n)`.
* `copies.duplicatecopies()` was renamed to
`copies.graftcopies()`. Its arguments changed from revision numbers
to context objects. It also lost its `repo` and `skip` arguments
(they should no longer be needed).