In addition to letting the FDM do the low-level signal watching, this
patch also fixes a bug; multiple SIGCHLDs, be it delivered either through a
signal, or via a signalfd, can be coalesced, like all signals.
This means we need to loop on waitpid() with WNOHANG until there are
no more processes to reap.
This in turn requires a small change to the way reaper callbacks are
implemented.
Previously, the callback was allowed to do the wait(). This was
signalled back to the reaper through the callback’s return value.
Now, since we’ve already wait():ed, the process’ exit status is passed
as an argument to the reaper callback.
The callback for the client application has been updated accordingly;
it sets a flag in the terminal struct, telling term_destroy() that the
process has already been wait():ed on, and also stores the exit
status.
Add fdm_signal_add() and fdm_signal_del(). Signals added to the fdm
will be monitored, and the provided callback called as “soon as
possible” from the main context (i.e not from the signal handler
context).
Monitored signals are *blocked* by default. We use epoll_pwait() to
unblock them while we’re polling. This allows us to do race-free
signal detection.
We use a single handler for all monitored signals; the handler simply
updates the signal’s slot in a global array (sized to fit SIGRTMAX
signals).
When epoll_pwait() returns EINTR, we loop the global array. The
callback associated with each signal that fired is called.
In many places we have the following pattern:
tll_foreach(list, it)
free(it->item.thing);
tll_free(list);
Since all tll functions are macros, and thus inlined, and since
tll_free in itself expands to a tll_foreach(), the above pattern
expands to more native code than necessary.
This is somewhat smaller:
tll_foreach(list, it) {
free(it->item.thing);
tll_remove(list, it);
}
Up until now, the various key binding modes (“normal”, “search” and
“url”) have used their own struct definitions for their key bindings.
The only reason for this was to have a properly typed “action” (using
the appropriate “action” enum).
This caused lots of duplicated code. This patch refactors this to use
a single struct definition for the “unparsed” key bindings handled by
the configuration, and another single definition for “parsed” bindings
used while handling input.
This allows us to implement configuration parsing, keymap translation
and so on using one set of functions, regardless of key binding mode.
* Use a do..while loop; this lets us drop the second half of the loop
condition.
* Call wcslen(prefix) once, *before* iterating the alphabet
characters.
* Step through the alphabet characters using a pointer, as this
avoids an indexed load (with possibly an imul instruction in
e.g. -Os builds).
* We were using the ‘back’ element of the list as prefix for the next
iteration of sequences, instead of the element at index ‘offset’
* ALEN() on a wchar_t[] includes the NULL terminator. We don’t want
that.
