On 07/20/10 14:47, Eric Blake wrote a bunch of good questions,
ending with:
> And how does signal handling fit in with interrupting a child test?
OK, you're right, a plain pipe is not going to work. How about
this much-smaller fix instead? It builds on your idea to make
sure that the 'read at_token' commands succeed.
--- general.m4 2010-07-20 14:28:32.835139154 -0700
+++ /tmp/general.m4 2010-07-20 14:48:26.295135987 -0700
@@ -1386,7 +1386,7 @@ dnl optimize away the _AT_CHECK subshell
dnl Ignore PIPE signals that stem from writing back the token.
trap "" PIPE
echo stop > "$at_stop_file"
- echo token >&AT_JOB_FIFO_FD
+ echo >&AT_JOB_FIFO_FD
dnl Do not reraise the default PIPE handler.
dnl It wreaks havoc with ksh, see above.
dnl trap - 13
@@ -1401,7 +1401,7 @@ dnl kill -13 $$
at_failed=:
fi
at_fn_group_postprocess
- echo token >&AT_JOB_FIFO_FD
+ echo >&AT_JOB_FIFO_FD
) &
$at_job_control_off
if $at_first; then
@@ -1410,7 +1410,9 @@ dnl kill -13 $$
fi
shift # Consume one token.
if test $...@%:@] -gt 0; then :; else
- read at_token <&AT_JOB_FIFO_FD || break
+ # The 'until' loop avoids a race condition if we happen to open
+ # the fifo between the time a child outputs a token and it exits.
+ until read at_token; do :; done <&AT_JOB_FIFO_FD
set x $[*]
fi
test -f "$at_stop_file" && break
@@ -1422,7 +1424,8 @@ dnl kill -13 $$
shift
for at_job
do
- read at_token
+ # Another 'until' loop to avoid the same race condition.
+ until read at_token; do :; done
done <&AT_JOB_FIFO_FD
fi
exec AT_JOB_FIFO_FD<&-
