A terminal with lots of scrollback history will have allocated a lot
of memory.
Normally, free() wont return this memory to the OS, and we don't seem
to trigger the automatic trim calls.
This means the server would accumulate quite a lot of memory over
time, as terminals come and go.
Now we explicitly trim the memory every time a terminal is destroyed.
Make ptmx non-blocking. Then, when writing data to the slave would
have blocked, use the FDM to asynchronously write the remaining data.
This is done by enabling EPOLLOUT on ptmx, and enqueueing all outgoing
data. The FDM handler will go through the enqueued data, and once all
of it has been written, we turn off EPOLLOUT again (thus switching
back to synchronous writes)
While unusual, it *is* possible for a client *not* to terminate when
we close ptmx.
We need to handle this *somehow*. Since it is so unusual, we'll go
with a fairly easy, but synchronous method:
* Register a signal handler for SIGALRM, and setup a 2 second alarm
* Wait for slave to die
* If it didn't die, sent SIGTERM, then re-set the alarm for another 2
seconds.
* If it still hasn't died, send SIGKILL (this time without an alarm).
When there hasn't been a timeout (or in our case, there was a timeout,
but we reset the timer before we got to the read()), read() returns,
not 0, but -1 with errno == EAGAIN.
Since we cancel the timers every now and then, there's a (small)
chance that one handler cancels a timer that has triggered in the same
epoll() iteration.
When this happens, read() blocks.
Fix by making the timer FDs non-blocking, and simply returning when we
read 0 bytes.
This function unmaps the terminal window, removes itself from the
wayland list of terminals and then finally destroys itself.
We ensure we don't get any callbacks/events referring to a free:d
terminal struct, we close all terminal related FDs and unmap the
wayland window.
Then, to be really really sure there aren't any (by the FDM) queued up
events, we delay the self-destruction to the next FDM poll iteration,
by opening an event FD and adding it to the FDM.
The callback for the event FD removes the FD from the FDM again, and
closes it. And then proceeds to destroy the terminal.
Since we now initialize the worker threads from term_init(), which
returns before the threads terminate, we can no longer use
stack-allocated worker contexts.
We _could_ put them in the terminal struct. But a simpler solution is
to allocate them in term_init(), and let the threads free them when
they don't need them anymore.
We do however need access to it, so provide a pointer. The difference
is that now we can have a *single* wayland instance, but multiple
terminal instances.
With this assumption, we can replace 'a % b' with 'a & (b - 1)'. In
terms of instructions, this means a fast 'and' instead of a slow
'div'.
Further optimize scrolling by:
* not double-initializing empty rows. Previously, grid_row_alloc()
called calloc(), which was then followed by a memset() when
scrolling. This is of course unnecessary.
* Don't loop the entire set of visible rows (this was done to ensure
all visible rows had been allocated, and to prefetch the cell
contents).
This isn't necessary; only newly pulled in rows can be NULL. For
now, don't prefetch at all.
That is, remove the 'regular' and 'bright' color arrays. This is
possible since the 256-color array is defined such that the first 16
colors map to the regular and bright colors.
If we scroll enough, we'll eventually end up wrapping around the
entire scrollback buffer. At this point, a selection is no longer
valid, so cancel it.
Note: this was very obvious when scrolling in the alt screen, since
its scrollback buffer is what you see on the screen (i.e. it has no
scrollback).
