`waitForEvent dpy 0` eventually waits indefinitely about x11

[geekosaur]

It looks like the original author of the X11 package assumed struct timeval contains 32-bit quantities, but they're actually 64 bits. This means the upper bits can contain garbage.

[geekosaur]

Notably, it looks to me like the lower 32 bits of tv_sec are 0 as expected, but the upper 32 bits aren't.

[geekosaur]

You'll want to depend on X11 ^>= 1.10.3.10, although that version isn't on Hackage yet so you'll have to vendor it for now.

[geekosaur]

The complication here is we may still need to support platforms where tv_sec is time_t (32 bits). Not sure how we deal with this, especially when we can't rely on it using glibc (we still have NetBSD users, I think). Probably something in hsc2hs?

[liskin]

There's probably something we can #if on?
(too sick to open a computer so not looking into it any further now, sorry)

[geekosaur]

I'm not finding anything portable, aside from doing it in configure. But, it seems the unix package assumes it's CLong, so we might as well do so too.

[geekosaur]

Okay, in the interests of not forcing a PVP break, I left waitForEvent with the same type; but it should at some point change to either Word64 or CLong.

[slotThe]

We are following PVP? :)