I just realized that there was another use of this field as sync lock, which I changed to use 'this' too. The lucidaFontName field is used in one other location (not for syncing) and it surely could use some cleanup, but I left it alone for now.
http://cr.openjdk.java.net/~rkennke/6892485/webrev.01/ I need at least another review (not sure if Igor needs to say 'go' again?). Thanks, Roman Am Donnerstag, den 25.02.2010, 11:21 -0800 schrieb Igor Nekrestyanov: > i am ok with this fix. > > -igor > > On 2/25/10 7:19 AM, Roman Kennke wrote: > > Hi there, > > > > this patch fixes the deadlock in FontManager for OpenJDK7 that has been > > reported repeatedly on this list. I have a testcase here (attached), > > which I could prepare for jtreg, but it is not totally reliable and > > depends on the system configuration (which and how many fonts > > installed), not sure if we want this. > > > > The problem is that two threads run into font code, locking on two > > different locks in wrong order. (Better analysis is in the bug). The fix > > is to synchronize only on one lock object (the FontManager instance). > > > > http://cr.openjdk.java.net/~rkennke/6892485/webrev.00/ > > > > Ok to commit? > > > > >
