Hi,
I fixed the bug 6938627 but I'm not sure if it should have been fixed,
or rather it causes other problems with the jps and other tools which
expect /tmp, so I have to go back and see if it was correct and/or fix
the jps/jstat tools. I wouldn't expect the VM to delete the directory
as you said in either case.
I'm currently tracking this as bugs:
6447220: HotSpot and JDK don't agree on /tmp location on Solaris
6447182: temp dir locations should not be hardcoded for
hsperfdata_<USER> dirctories
Thanks,
Coleen
On 11/09/10 14:20, Phil Race wrote:
Pavel,
The font implementation and tests are 2D. Not AWT. Don't get thrown by
package names.
I was a bit curious about the real reason for this failure.
I wrote this test a couple of years ago and its never failed before.
I don't see it failing on any 6-open build.
I see it failing only in JDK 7 from b90. That seems to be due to
HS18b03 fix
6938627: Make temporary directory use property java.io.tmpdir when
specified
I suppose the VM doesn't delete the directory on exit even if empty
because it
could interfere with other VM instances.
I'm not sure what more the test can do than try to deal with any
artifacts
created by some other part of the JRE. As it is, its making sure that
the tmp
dir is private to its own private JRE invocation. But that doesn't
mean the
test and the functionality its testing get exclusive rights to this
directory.
So I don't think its worth trying to suppress the VM behaviour either.
If the VM or something else changes the name of this file or adds
another one,
then that's going to require fix up such as this one. Just have to
hope its
not a regular occurrence.
So I'm OK with the fix.
-phil.
On 11/9/2010 7:40 AM, Pavel Tisnovsky wrote:
Hi all,
could anybody please review correction of
awt/FontClass/CreateFont/DeleteFont regression test?
This test check, if the temporary file containing font is properly
deleted when JVM finishes. The check is based on counting of files in
temporary dir - if temporary font file is not deleted, the number of
files before and after running the test differs.
BUT JVM also creates hsperfdata_* subdirectory so the check described
above always fails (presence of this subdirectory add +1 to number of
files). I've added simple filter to resolve this issue. Another option
is to disable the hsperfdata_* creation at all.
Webrev is available at:
http://cr.openjdk.java.net/~ptisnovs/DeleteFont_correction/
PS: I'd also like to backport this correction to JDK6 if it is possible.
Thank you in advance
Pavel T.
