Just a quick check, jdk.accessibility is only linked in windows image at the moment. It is a bug. Are you going to fix that in this changeset? I think you have to verify this change in windows as well as other platforms.
Mandy > On Jun 10, 2015, at 3:33 PM, Pete Brunet <peter.bru...@oracle.com> wrote: > > Due to some other priorities it's been over 2 months since the last webrev. > An update is here: > http://cr.openjdk.java.net/~ptbrunet/JDK-8055160/webrev.03 > > The changes from webrev.02 are: > > 1) The test was changed to not use the service provider to test the > activation of the service provider. Instead a file is created when > Toolkit.getDefaultToolkit activates providers and tested for existence when > the test runs. > > 2) The copyright header in the new jdk.accessibility files were fixed. > > Pete > > On 4/3/15 3:59 PM, Pete Brunet wrote: >> Due to the recent push of JDK-8076182 (Open source Java Access Bridge) which >> exposed some files that were in closed the webrev needs a full re-review. >> I've also made the changes requested by Mandy. >> >> http://cr.openjdk.java.net/~ptbrunet/JDK-8055160/webrev.02/ >> >> Pete >> >> On 3/23/15 4:41 PM, Mandy Chung wrote: >>> >>> >>> On 3/19/2015 6:03 PM, Pete Brunet wrote: >>>> A new webrev is available at >>>> http://cr.openjdk.java.net/~ptbrunet/JDK-8055160/webrev.01/ >>>> >>> >>> line 820-821: this comment is incorrect. >>> >>> line 831-838: what happens if ServiceConfigurationException thrown or any >>> exception is thrown by the activate method? This should wrap with AWTError >>> as I mentioned in my previous review comment. This was hidden with the >>> test (see below). >>> >>> line 891-901: this example may not be necessary as the service loader >>> documentation should cover it. >>> >>>> >>>> >>>> The changes to the tests are: >>>> - added an unused provider >>>> - added a test activating two providers >>>> >>>> Mandy, Regarding the last bullet I'm not sure I resolved your comment, >>>> "For the test, since you support multiple providers, perhaps good to add >>>> one more test case to activate two providers and load two providers but >>>> only one is activated." If not, please let me know. >>> >>> Almost. For Foo, Bar providers, their activate method throwing >>> RuntimeException actually stops loading the second provider. The activate >>> method could perhaps update some static field defined in the Load class if >>> it's called (perhaps adding its name) so that you can tell whether the >>> expected providers are activated. UnusedProvider throwing RuntimeException >>> is good since you don't expect it's activated. >>> >>> Otherwise, looks good. >>> >>> Mandy >> >