Re: ksh is broken

[christos%zoulas.com@localhost (Christos Zoulas)]

On Mar 17,  8:21pm, okuyama%flex.phys.tohoku.ac.jp@localhost (Rin Okuyama) wrote:
-- Subject: ksh is broken

| Hi,
| 
| ksh is broken due to changes made in bin/ksh/siglist.sh Rev 1.10:
| 
|   % ksh
|   $ /bin/echo a
|   a
|   (then, ksh is permanently trapped in __sigsuspend14)
| 
| This is because siglist.out, which is generated by siglist.sh, is not
| sorted, and SIGABRT = SIGIOT = 6 appears two times. The followings are
| codes from bin/ksh/trap.c, which apparently fails with siglist.out
| which is not sorted or has duplicated entries:
| 
| Trap sigtraps[SIGNALS+1] = {
|         { .signal = SIGEXIT_, .name = "EXIT", .mess = "Signal 0" },
| #include "siglist.out"  /* generated by siglist.sh */
|         { .signal = SIGERR_,  .name = "ERR",  .mess = "Error handler" },
|     };
| ...
|         sigtraps[SIGINT].flags |= TF_DFL_INTR | TF_TTY_INTR;
|         sigtraps[SIGQUIT].flags |= TF_DFL_INTR | TF_TTY_INTR;
|         sigtraps[SIGTERM].flags |= TF_DFL_INTR;/* not fatal for interactive */
|         sigtraps[SIGHUP].flags |= TF_FATAL;
|         sigtraps[SIGCHLD].flags |= TF_SHELL_USES;
| 
|         /* these are always caught so we can clean up any temporary files. */
|         setsig(&sigtraps[SIGINT], trapsig, SS_RESTORE_ORIG);
|         setsig(&sigtraps[SIGQUIT], trapsig, SS_RESTORE_ORIG);
|         setsig(&sigtraps[SIGTERM], trapsig, SS_RESTORE_ORIG);
|         setsig(&sigtraps[SIGHUP], trapsig, SS_RESTORE_ORIG);
| ...
| 
| I guess that siglist.sh Rev 1.10 is a work around for gcc 5, whose cpp
| generates noisy output. Here I attached another work around based on
| siglist.sh Rev 1.9. It is kind of an ugly hack, but at least works with
| both gcc 4.8 and 5.3.

Thanks!

christos