Use php_pollfd_for_ms more consistently for single fd polling

[NattyNarwhal]

There's a few cases where we call poll with a single fd; we have a wrapper (php_pollfd_for_ms) to make this case cleaner with less boilerplate.

Side note: said wrapper also uses the php_poll2 API, which can also emulate poll(2) using select(2); FastCGI was doing this on its own, so that also handles that case. That said, poll seems to be basically everywhere nowadays, so emulating poll is only seems useful for Windows, which has WSAPoll. I suspect php_poll2 should be a WSAPoll wrapper instead? The new poll API handles all this of course, but seems pretty heavyweight for the case of just quickly polling a single fd.

[NattyNarwhal]

There's a few cases where we call poll with a single fd; we have a wrapper (php_pollfd_for_ms) to make this case cleaner with less boilerplate.

Side note: said wrapper also uses the php_poll2 API, which can also emulate poll(2) using select(2); FastCGI was doing this on its own, so that also handles that case. That said, poll seems to be basically everywhere nowadays, so emulating poll is only seems useful for Windows, which has WSAPoll. I suspect php_poll2 should be a WSAPoll wrapper instead? The new poll API handles all this of course, but seems pretty heavyweight for the case of just quickly polling a single fd.

FWIW looked into this further and there are only a few places where select(2) is called that outside of fallbacks if poll(2) isn't present (vendored libmagic, mysqlnd for mysqli_poll, streams_select, socket_select, dev CLI web server, CLI for polling a single fd, FPM backend), and we have a big wrapper in win32/ for reimplementing select on Windows. Not sure how useful it'd be to make those all consistent with poll or using the polling API and get rid of the reimplementations.

[arnaud-lb]

Looks good to me otherwise

[NattyNarwhal]

Thanks, I'll take a further look at the poll/select stuff I mentioned in this thread soon enough.