Mercurial > emacs
changeset 18405:4eaaff49a24e
(read_avail_input): Check for failure with EIO.
| author | Richard M. Stallman <rms@gnu.org> |
|---|---|
| date | Mon, 23 Jun 1997 02:39:44 +0000 |
| parents | 2af2dc0cd97f |
| children | 5c2e5ccd0acb |
| files | src/keyboard.c |
| diffstat | 1 files changed, 7 insertions(+), 0 deletions(-) [+] |
line wrap: on
line diff
--- a/src/keyboard.c Mon Jun 23 02:30:36 1997 +0000 +++ b/src/keyboard.c Mon Jun 23 02:39:44 1997 +0000 @@ -4924,6 +4924,13 @@ #else nread = read (input_fd, cbuf, n_to_read); #endif + /* POSIX infers that processes which are not in the session leader's + process group won't get SIGHUP's at logout time. BSDI adheres to + this part standard and returns -1 from read(0) with errno==EIO + when the control tty is taken away. + Jeffrey Honig <jch@bsdi.com> says this is generally safe. */ + if (nread == -1 && errno == EIO) + kill (0, SIGHUP); #if defined (AIX) && (! defined (aix386) && defined (_BSD)) /* The kernel sometimes fails to deliver SIGHUP for ptys. This looks incorrect, but it isn't, because _BSD causes