I was learning from @jdelong and he said that you should use
double quotes for local includes and angle brackets for library
includes. I asked why our code was the way it was, and he said he wanted
to clean it up. I beat him to it :)
Conflicts:
hphp/runtime/base/server/admin_request_handler.cpp
hphp/runtime/vm/named_entity.h
I think the cause of our issues with random failed-to-exec
errors is this SIGCHLD handler. When the parent process asks a light
process to do a waitpid operation, we now can get EINTR for two
reasons. It used to use signals only for implementing timeouts with
SIGALRM, but now it appears possible due to the SIGCHLD. This could
lead us to report the waitpid as failing---which it looks can have
ExecFuture thinking running==false and reading the exit code as -1.
Also, now that we have a SIGCHLD handler (are we sure we want to keep
this?), using waitpid() without an EINTR loop is probably similarly
broken in the parent process. (Also, isn't it basically incorrect to
use wait functions without an EINTR loop in general, though?) I
didn't add an EINTR loop in do_waitpid in this diff because it's using
signals for timeout (it really should be checking a volatile
sigatomic_t that gets set by the handler, though)
Also, this means using Process::Exec is suspect in general. All uses
looked like debugger or tests only (and when hphpc runs hhvm in a
subprocess), so I didn't nuke them (yet). I deleted some random
network.h dead code that uses it, though.
This change is mostly for FB internal organizational reasons.
Building is not effected beyond the fact that the target now
lands in hphp/hhvm/hhvm rather than src/hhvm/hhvm.
Tim Starling