On my system, aclocal7.test would spuriously fail about 1 out of 3
tests. I tracked that down to an `$AUTOMAKE' creating `Makefile.in'
with a time stamp of the same second as the file `foo' being touched
right before. This confuses the age comparison done by `ls -1t'.
Addition of the second `$sleep' in the patch below fixes that. I added
two more for the eventual system where both aclocal and automake finish
before a new second (only a matter of time, and I think 1 second
granularity file systems aren't dead by then).
OK to apply?
Cheers,
Ralf
* tests/aclocal7.test: Add `$sleep's between file touching
and automake resp aclocal+automake invocations, to ensure
they complete with a time stamp strictly later than the touched
file.
Index: tests/aclocal7.test
===================================================================
RCS file: /cvs/automake/automake/tests/aclocal7.test,v
retrieving revision 1.3
diff -u -r1.3 aclocal7.test
--- tests/aclocal7.test 14 May 2005 20:28:53 -0000 1.3
+++ tests/aclocal7.test 22 Apr 2006 19:36:14 -0000
@@ -89,6 +89,7 @@
touch fragment.inc
$sleep
touch foo
+$sleep
$ACLOCAL -I m4
$AUTOMAKE --no-force
# Only ./Makefile.in should change.
@@ -101,6 +102,7 @@
: > README
$sleep
touch foo
+$sleep
$AUTOMAKE --no-force
# Even if no dependency change, the content changed.
test `ls -1t Makefile.in foo | sed 1q` = Makefile.in
@@ -111,6 +113,7 @@
: > sub/Makefile.in
$sleep
touch foo
+$sleep
$ACLOCAL -I m4
$AUTOMAKE --no-force
# Only sub/Makefile.in should change.
