I don't understand why, but I had to make a small change in order to get
control+d to work properly with Python + WASI. For some reason
e->el_tty.t_c[TS_IO] is just all zeros, whereas e->el_tty.t_c[EX_IO] is
interesting. Reading all the code (!), it makes way more sense to me
to consider e->el_tty.t_c[EX_IO] rather than e->el_tty.t_c[TS_IO].
Maybe this is just a decades old bug in libedit, and people usually use
readline, so they don't notice. I don't know.
I found that using e->el_tty.t_c[EX_IO][C_EOF] led to a lot of problems in
various states as well after revamping our IO and refactoring code. I'm
now just hardcoding the number 4, which seems to work very well in all cases for us.
--- readline-orig.c 2022-07-20 16:28:37.000000000 -0700
+++ readline.c 2022-07-20 16:49:05.000000000 -0700
@@ -2112,7 +2112,7 @@
el_set(e, EL_UNBUFFERED, 1);
if (buf == NULL || count-- <= 0)
return;
- if (count == 0 && buf[0] == e->el_tty.t_c[TS_IO][C_EOF])
+ if (count == 0 && buf[0] == 4)
done = 1;
if (buf[count] == '\n' || buf[count] == '\r')
done = 2;
