harmony-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Mark Hindess <mark.hind...@googlemail.com>
Subject Re: Remaining M9 issues
Date Wed, 25 Mar 2009 19:53:42 GMT

In message <3301EDEF-1A1F-423D-9425-7B9DA200215A@gmail.com>, Nathan Beyer
> On Mar 25, 2009, at 5:40 AM, Tim Ellison <t.p.ellison@gmail.com> wrote:
> > Nathan Beyer wrote:
> >> On Tue, Mar 24, 2009 at 9:10 AM, Tim Ellison  
> >> <t.p.ellison@gmail.com> wrote:
> >>> 3. Various AWT/Swing crashes/ failures
> >>
> >> Linux AWT/Swing crashes - works on Ubuntu 8.04, doesn't on 8.10
> >> https://issues.apache.org/jira/browse/HARMONY-6123
> >
> > Looks like a problem in the X.11 libraries, and looking around it  
> > seems that we are far from the only ones who have uncovered this problem.
> >
> Agreed, but I got the impression that the new behavior is considered  
> correct and it will be up to apps to fix their own code. Have you  
> found any indication to the contrary?

Well, there have been mistakes in the past - such as nested locking not
working as documented in the XLockDisplay man page and with locks being
held by the toolkit when user callbacks were invoked - but I think it is
quite likely that we need to do something to fix this.

We currently call XInitThreads to allow multi-threaded support but we
do not actually seem to call XLockDisplay.  I don't think this can be
correct though I don't know which xlib calls need to be locking - I'm
guessing at least the two mentioned in your JIRA.  If we had access
to the native bridge tool, we could probably make a safe fix of just
wrapping all Xlib calls with a lock but that would probably be overkill.
Having access to the native bridge tool would be useful anyway.[0]

I will try to get an ubuntu 8.10 kvm image installed and have a look at this 
problem but since awt/swing are not really a priority for me.

> > Since it is not a regression, and apparently not caused by anything we
> > can fix (unless we can find a workaround), I propose that we declare  
> > it a "Non-Blocker" for M9.


> I'm fine with that. Maybe indicate it is a known issue.

1+  It should be mentioned in the release notes.


[0] I think there are some modularity problems lurking here... I think
luni depends on the org.apache.harmony.misc.accessors which depends on
some native bridge code in awt.  I suspect the common native bridge
probably should be brought back in to the misc module.

View raw message