Re: while-no-input

[Kim F. Storm]

Richard Stallman <address@hidden> writes:

> So here's my simpler proposal for how to implement this.

I tried it -- but I don't have SYNC_INPUT defined, so I used this:

#define QUIT                                            \
  do {                                                  \
    if (!NILP (Vquit_flag) && NILP (Vinhibit_quit))     \
      {                                                 \
        Lisp_Object flag = Vquit_flag;                  \
        Vquit_flag = Qnil;                              \
        if (EQ (Vthrow_on_input, flag))                 \
          Fthrow (Vthrow_on_input, Qnil);               \
        Fsignal (Qquit, Qnil);                          \
      }                                                 \
  } while (0)


I then ran this code (in my normal emacs with cua-mode enabled):

(while-no-input
  (while t
    (prin1 "x")
    (sit-for 0)))

and hit space after a few seconds.  Got a crash:


Breakpoint 1, abort () at emacs.c:442

(gdb) xbacktrace
"cua--post-command-handler"
"run-hooks"
(gdb) bt
#0  abort () at emacs.c:442
#1  0x08182ae9 in Feval (form=140121845) at eval.c:2015
#2  0x081819e0 in Fcondition_case (args=145132605) at eval.c:1329
#3  0x081b7204 in Fbyte_code (bytestr=144710763, vector=144806556, maxdepth=24) 
at bytecode.c:864
#4  0x08184383 in funcall_lambda (fun=144661804, nargs=0, 
arg_vector=0xbfffd608) at eval.c:2959
#5  0x08183dcb in Ffuncall (nargs=1, args=0xbfffd604) at eval.c:2820
#6  0x081836c8 in run_hook_with_args (nargs=1, args=0xbfffd604, 
cond=to_completion) at eval.c:2440
#7  0x081834bc in Frun_hooks (nargs=1, args=0xbfffd6d4) at eval.c:2308
#8  0x08183b46 in Ffuncall (nargs=2, args=0xbfffd6d0) at eval.c:2753
#9  0x0818386b in call1 (fn=137798905, arg1=137754665) at eval.c:2566
#10 0x0810cf01 in safe_run_hooks_1 (hook=0) at keyboard.c:2034
#11 0x08181b06 in internal_condition_case (bfun=0x810cee7 <safe_run_hooks_1>, 
handlers=137698537, hfun=0x810cf06 <safe_run_hooks_error>) at eval.c:1382
#12 0x0810cf9c in safe_run_hooks (hook=137754665) at keyboard.c:2062
#13 0x0810c25b in command_loop_1 () at keyboard.c:1801
#14 0x08181b06 in internal_condition_case (bfun=0x810adbb <command_loop_1>, 
handlers=137764305, hfun=0x810a900 <cmd_error>) at eval.c:1382
#15 0x0810ac33 in command_loop_2 () at keyboard.c:1309
#16 0x081815cc in internal_catch (tag=137756753, func=0x810ac14 
<command_loop_2>, arg=137698489) at eval.c:1143
#17 0x0810abe6 in command_loop () at keyboard.c:1288
#18 0x0810a67f in recursive_edit_1 () at keyboard.c:981
#19 0x0810a7c0 in Frecursive_edit () at keyboard.c:1042
#20 0x0810911b in main (argc=3, argv=0xbfffdf04) at emacs.c:1742
#21 0x42015574 in __libc_start_main () from /lib/tls/libc.so.6
(gdb) p handling_signal
$1 = 1
(gdb) p interrupt_input_blocked
$2 = 0
(gdb) 

Seems that we can break out of XTread_socket without restoring
handling_signal.

I don't know if it is sufficient to unwind protect it.


-- 
Kim F. Storm <address@hidden> http://www.cua.dk