lock-and-run.sh: Check for process existence rather than timeout.

	* lock-and-run.sh: Check for process existence rather than timeout.

Matthias Klose noted that on less powerful targets, a link might take more
than 5 minutes; he mentions a figure of 3 hours for an LTO link.  So this
patch changes the timeout to a check for whether the locking process still
exists.  If the lock exists in an erroneous state (no pid file or can't
signal the pid) for 30 sec, steal it.

From-SVN: r277277

diff --git a/gcc/ChangeLog b/gcc/ChangeLog
index 2aa5536cf329fca1129c3c690cd584049ad1abee..5e6c6620d17364cb468f08463d3cbd770c0a2ad6 100644 (file)
--- a/gcc/ChangeLog
+++ b/gcc/ChangeLog
@@ -1,3 +1,7 @@
+2019-10-21  Jason Merrill  <jason@redhat.com>
+
+       * lock-and-run.sh: Check for process existence rather than timeout.
+
 2019-10-21  Jozef Lawrynowicz  <jozef.l@mittosystems.com>
 
        * expr.c (expand_expr_real_2): Don't widen constant op1 when expanding

diff --git a/gcc/lock-and-run.sh b/gcc/lock-and-run.sh
index 3a6a84c253ae1f2fcdea7ccfb586b0796e3199d5..22bc436565cc0dee76202a3248cdc3ad253642a1 100644 (file)
--- a/gcc/lock-and-run.sh
+++ b/gcc/lock-and-run.sh
@@ -1,33 +1,49 @@
 #! /bin/sh
-# Shell-based mutex using mkdir.
+# Shell-based mutex using mkdir.  This script is used in make to prefer
+# serialized execution to avoid consuming too much RAM.  If reusing it,
+# bear in mind that the lock-breaking logic is not race-free, so disable
+# it in err() if concurrent execution could cause more serious problems.
 
+self=`basename $0`
 lockdir="$1" prog="$2"; shift 2 || exit 1
 
 # Remember when we started trying to acquire the lock.
 count=0
-touch lock-stamp.$$
 
-trap 'rm -r "$lockdir" lock-stamp.$$' 0
+err () {
+    if test -f $lockdir/lock-$1.$$; then
+       echo "$self: *** (PID $$) removing stale $lockdir" >&2
+       rm -rf $lockdir
+       # Possible variant for uses where races are more problematic:
+       #echo "$self: *** (PID $$) giving up, maybe rm -r $lockdir" >&2
+       #exit 42
+    else
+       touch $lockdir/lock-$1.$$
+    fi
+}
 
 until mkdir "$lockdir" 2>/dev/null; do
     # Say something periodically so the user knows what's up.
     if [ `expr $count % 30` = 0 ]; then
-       # Reset if the lock has been renewed.
-       if [ -n "`find \"$lockdir\" -newer lock-stamp.$$`" ]; then
-           touch lock-stamp.$$
-           count=1
-       # Steal the lock after 5 minutes.
-       elif [ $count = 300 ]; then
-           echo removing stale $lockdir >&2
-           rm -r "$lockdir"
+       # Check for valid lock.
+       if pid=`cat $lockdir/pid 2>/dev/null` && kill -0 $pid 2>/dev/null; then
+           echo "$self: (PID $$) waiting $count sec to acquire $lockdir from PID $pid" >&2
+       elif test -z "$pid"; then
+           echo "$self: (PID $$) cannot read $lockdir/pid" >&2
+           err nopid
        else
-           echo waiting to acquire $lockdir >&2
+           echo "$self: (PID $$) cannot signal $lockdir owner PID $pid" >&2
+           err dead
        fi
     fi
     sleep 1
     count=`expr $count + 1`
 done
 
+trap 'rm -rf "$lockdir"' 0
+echo $$ > $lockdir/pidT && mv $lockdir/pidT $lockdir/pid
+echo "$self: (PID $$) acquired $lockdir after $count seconds" >&2
+
 echo $prog "$@"
 $prog "$@"