Re: HELP! [un]request_sigio_on_device() fubar'd on Linux

Ben Wing <ben(a)666.com&gt; writes: Yes I have but only on Linux 2.2/glibc2.2 so far. I am worried that this might break on random system. If it is really only ULTRIX. There is also a comment referring to "some kernels", whatever that means. I will test removing the slow_down_interrupts/speed_up_interrupts on some more platforms I have access to and then apply a patch to the trunk. We should then encourage people to test this in the next beta. Martin (or anybody else), maybe you could also try this on your amazing collection of random machines :-) Please try the following: - apply the sysdep.c patch - start xemacs -vanilla and eval (open-network-stream "bug" nil "10.0.0.42" 42) - press some random keys (to cause SIGIOs and see whether connect survies restarting) - try to interrupt with C-g - wait for upto 15 secs and see whether C-g worked - now apply the process-unix.c patch and try again. If C-g doesn't work after applying the patches then the output of strace/truss might be useful. thanks in advance! Gunnar P.S.: Of course we should also fix x_check_for_quit_char(). I will send a separate message about that. Index: sysdep.c =================================================================== RCS file: /usr/CVSroot/XEmacs/xemacs/src/sysdep.c,v retrieving revision 1.38 diff -u -r1.38 sysdep.c --- sysdep.c 2001/04/12 18:24:22 1.38 +++ sysdep.c 2001/04/16 12:20:17 ＠＠ -1108,7 +1108,7 ＠＠ { int filedesc = DEVICE_INFD (d); -#if defined (I_SETSIG) && !defined(HPUX10) +#if defined (I_SETSIG) && !defined(HPUX10) && !defined(LINUX) { int events=0; ioctl (filedesc, I_GETSIG, &events); Index: process-unix.c =================================================================== RCS file: /usr/CVSroot/XEmacs/xemacs/src/process-unix.c,v retrieving revision 1.20 diff -u -r1.20 process-unix.c --- process-unix.c 2001/04/12 18:24:11 1.20 +++ process-unix.c 2001/04/16 12:00:20 ＠＠ -1721,8 +1721,9 ＠＠ /* Slow down polling. Some kernels have a bug which causes retrying connect to fail after a connect. */ +#if 0 slow_down_interrupts (); - +#endif loop: /* A system call interrupted with a SIGALRM or SIGIO comes back ＠＠ -1730,7 +1731,9 ＠＠ SETJMP (break_system_call_jump); if (QUITP) { +#if 0 speed_up_interrupts (); +#endif REALLY_QUIT; /* In case something really weird happens ... */ slow_down_interrupts (); ＠＠ -1764,9 +1767,11 ＠＠ failed_connect = 1; close (s); + s = -1; +#if 0 speed_up_interrupts (); - +#endif continue; } ＠＠ -1791,7 +1796,9 ＠＠ break; } /* address loop */ +#if 0 speed_up_interrupts (); +#endif freeaddrinfo (res); if (s < 0) ＠＠ -1852,7 +1859,9 ＠＠ /* Slow down polling. Some kernels have a bug which causes retrying connect to fail after a connect. */ +#if 0 slow_down_interrupts (); +#endif loop: ＠＠ -1861,7 +1870,9 ＠＠ SETJMP (break_system_call_jump); if (QUITP) { +#if 0 speed_up_interrupts (); +#endif REALLY_QUIT; /* In case something really weird happens ... */ slow_down_interrupts (); ＠＠ -1892,13 +1903,17 ＠＠ close (s); +#if 0 speed_up_interrupts (); +#endif errno = xerrno; report_file_error ("connection failed", list2 (host, name)); } +#if 0 speed_up_interrupts (); +#endif #endif /* ! HAVE_GETADDRINFO */ }