Hi Petr,
Hello, Kumar.
Thank you for the review.
Is there a regression test for this somewhere ?
It's quite hard to create a regression test for this issue, as we would need
SWT for such test and we do not have any tests which use SWT currently.
I have tried to make a test without SWT using some apple's APIs to check we do
not block the queue, but it does not look like there's an easy way to do it.
In that case we need to mark JDK-8009911 with label "noreg-hard", with
your explanation
as to why a test is infeasible.
So, are you OK with the fix without the test?
Yes go for it.
Kumar
With best regards. Petr.
On May 24, 2013, at 4:51 AM, Kumar Srinivasan wrote:
Hi Petr,
Sorry for the delay, I somehow missed this one.
I don't understand objective-c, if Mr Petrov is happy I am ok with it.
There seems to be indenting issues 1049-1051
Is there a regression test for this somewhere ?
Thanks
Kumar
Hello,
Yes; Kumar should be one of the reviewers here.
Thanks,
-Joe
On 05/21/2013 04:18 AM, Anthony Petrov wrote:
Hi Petr,
The fix looks good to me. We still need a review from someone who maintains the
Launcher code. Kumar?
Also, you could use the @autoreleasepool directive to simplify it a bit,
although I'm fine with the explicit syntax, too.
One comment regarding your remark:
c. We would have lost all the hg history for this file.
We wouldn't. Mercurial handles renaming of files very well. See `hg help
rename` for more details. (I'm not suggesting to rename this file though.)
--
best regards,
Anthony
On 05/21/2013 10:57 AM, David Holmes wrote:
Adding core-libs as this is actually a launcher change.
David
On 21/05/2013 4:36 PM, Petr Pchelko wrote:
Hello, AWT Team.
Please review the fix for the issue:
http://bugs.sun.com/view_bug.do?bug_id=8009911
The webrev is available at:
http://cr.openjdk.java.net/~pchelko/8009911/webrev.00/
The problem:
When launching java with -XstatrOnFirstThread we were using
dispatch_sync to start a Java main on the main thread. This blocked
the main dispatch queue, so all the subsequent calls to GCD were never
invoked. GCD is used a bit in our code, but it is used inside Cocoa,
for example for fullscreen.
The solution:
We now launch Java main using a performSelectorOnMaintThread, so we do
not block GCD.
Notes:
1. This also fixes the following SWT bug:
https://bugs.eclipse.org/bugs/show_bug.cgi?id=388886
2. Although I have added OBJC syntax into java_md_macosx.c file, I did
not rename it to .m, because:
a. It is hard to update an old build system to compile with the
new file name.
b. It is already compiled with -x objective-c, so I will compile
c. We would have lost all the hg history for this file.
With best regards. Petr.
