[Toybox] Android microcom questions.

[Rob Landley]

Microcom says it's from "the android open source project", so presumably I
should send questions that way: Why do you open the tty in nonblocking mode and
then immediately switch to blocking mode for the reads (guarded by poll so they
shouldn't block anyway)? Are there ttys that hang on open?

I remember testing whether a CDROM drive had a CD in it without waiting to spin
it up involves trying to open the block device in nonblock mode and seeing what
error message you get. Presumably you can pull a similar trick on external USB
drives that spin down if you don't want to wait for them. But I dunno how it
applies to char devices?

I'm combining xraw() with set_terminal(), and moving the speed setting logic
into set_terminal. And while we're at it:

  $ echo hello | toybox microcom -s 115200 /dev/ttyS0
  microcom: tcgetattr fd: I/O error

(I think about these things because "toybox expect" is on my post-1.0 todo
list.) And if xraw() fails on stdin we haven't set up the signal handler yet and
don't restore the tty state? Dunno if it matters...

Heh. If you run "toybox top | cat" and then ctrl-c to kill it, the cursor
display isn't reenabled and the _reason_ is it was disabled by an ANSI escape
sequence and the signal handler that re-enables it outputs the corresponding
ANSI escape sequence... which can't make it out through cat if the ctrl-c killed
the whole pipeline.

I'm calling that one pilot error, largely because I dunno how to fix it. (The
tcsetattr call takes it back out of raw mode, but the tcsetattr man page doesn't
contain the word "cursor"...)

Rob