summaryrefslogtreecommitdiff
path: root/tests/ChangeLog
diff options
context:
space:
mode:
Diffstat (limited to 'tests/ChangeLog')
-rw-r--r--tests/ChangeLog45
1 files changed, 38 insertions, 7 deletions
diff --git a/tests/ChangeLog b/tests/ChangeLog
index dd763b1..b40612f 100644
--- a/tests/ChangeLog
+++ b/tests/ChangeLog
@@ -1,14 +1,45 @@
+1999-09-16 Paul D. Smith <psmith@gnu.org>
+
+ * scripts/features/reinvoke: Remove invocations of "touch" in
+ makefiles. See the comments on the touch function rewrite below.
+ Note that UNIX touch behaves the same way if the file already
+ exists: it sets the time to the _local_ time. We don't want
+ this. This is probably a good tip for makefile writers in
+ general, actually... where practical.
+ * scripts/options/dash-l: Ditto.
+ * scripts/options/dash-n: Ditto.
+
+ * test_driver.pl (run_each_test): In retrospect, I don't like the
+ .lN/.bN/.dN postfix required by DOS. So, for non-DOS systems I
+ changed it back to use .log, .base, and .diff.
+
+ * run_make_tests.pl (set_more_defaults): Move the check for the
+ make pathname to here from set_defaults (that's too early since it
+ happens before the command line processing).
+ Create a new variable $port_type, calculated from $osname, to
+ specify what kind of system we're running on. We should integrate
+ the VOS stuff here, too.
+ (valid_option): Comment out the workdir/-work stuff so people
+ won't be fooled into thinking it works... someone needs to fix
+ this, though!
+
+ * scripts/functions/origin: Use $port_type instead of $osname.
+ * scripts/functions/foreach: Ditto.
+ * scripts/features/default_names: Ditto.
+
1999-09-15 Paul D. Smith <psmith@gnu.org>
* test_driver.pl (touch): Rewrite this function. Previously it
used to use utime() to hard-set the time based on the current
- local clock. This fails badly on networked filesystems where the
- FS server clock is skewed from the local clock: normally modifying
- a file causes it to get a mod time based on the _server's_ clock.
- Hard-setting it based on the _local_ clock causes gratuitous
- errors and makes the tests unreliable except on local filesystems.
- The new function will simply modify the file, allowing the
- filesystem to set the mod time as it sees fit.
+ local clock, or, if the file didn't exist, it merely created it.
+ This mirrors exactly what real UNIX touch does, but it fails badly
+ on networked filesystems where the FS server clock is skewed from
+ the local clock: normally modifying a file causes it to get a mod
+ time based on the _server's_ clock. Hard-setting it based on the
+ _local_ clock causes gratuitous errors and makes the tests
+ unreliable except on local filesystems. The new function will
+ simply modify the file, allowing the filesystem to set the mod
+ time as it sees fit.
* scripts/features/parallelism: The second test output could
change depending on how fast some scripts completed; use "sleep"