9 b58 fails on Win 8 when screen reader running

Pete Brunet peter.brunet at oracle.com
Thu Apr 23 02:26:42 UTC 2015


Hi Erik, It's https://bugs.openjdk.java.net/browse/INTJDK-7615864.

On 4/22/15 8:20 PM, Erik Joelsson wrote:
> That's an interesting find. I suppose we need to add that linker flag
> then. Do you have a bug opened for this issue?
>
> /Erik
>
> On 2015-04-22 17:56, Pete Brunet wrote:
>> The problem is fixed by doing this (using the VS2013 versions of the
>> dump/editbin):
>> editbin ...\java.exe /HIGHENTROPYVA:NO
>>
>> You can see the before and after using
>> dumpbin ...\java.exe /HEADERS
>>
>> Before:
>>
>> OPTIONAL HEADER VALUES
>> ...
>>              8160 DLL characteristics
>>                     High Entropy Virtual Addresses
>>                     Dynamic base
>>                     NX compatible
>>                     Terminal Server Aware
>>
>> After:
>>
>> OPTIONAL HEADER VALUES
>> ...
>>              8140 DLL characteristics
>>                     Dynamic base
>>                     NX compatible
>>                     Terminal Server Aware
>>
>> The following is from an industry contact who ran into the same thing in
>> their own app:
>>
>> The high-entropy 64-bit address space layout randomization feature is
>> specific to 64-bit versions of Windows 8. It is discussed in some detail
>> in the article "Software defense: mitigating common exploitation
>> techniques" found at
>> http://blogs.technet.com/b/srd/archive/2013/12/11/software-defense-mitigating-common-exploitation-techniques.aspx.
>>
>>
>> This feature is enabled by default in 64-bit executable images and can
>> be disabled via the /HIGHENTROPYVA:NO linker switch. See
>> http://msdn.microsoft.com/en-us/library/jj835761.aspx
>>
>> With the high-entropy 64-bit address space layout randomization option
>> enabled, all memory addresses in a 64-bit app are guaranteed to be above
>> 4gb. This means that if any module has latent pointer truncation issues
>> in 64-bit builds, these issues will manifest themselves when this flag
>> is set.
>>
>> Pete
>>
>> On 4/21/15 3:19 PM, Pete Brunet wrote:
>>> Hi Phil,
>>>
>>> On 4/21/15 2:42 PM, Phil Race wrote:
>>>> Perhaps JAWS is not well behaved w.r.t multiple DLL versions ?
>>> Tell me more.
>>>> Does it happen only on win 8 ? If you boot the same system into win 8
>>>> or win 8.1 is it OK ?
>>> Win 7 is no problem.  I run Win 8.1 Pro in a VirtualBox VM.  This bug
>>> was reported as
>>> https://bugs.openjdk.java.net/browse/intjdk-7615864
>>>> What do you see logged by hotspot from typing Ctrl-Break in the
>>>> shell ?
>>> It will be back to the command prompt before I can do that.
>>>> Other than that, debugging is very likely going to require you build
>>>> with VS2013
>>>>
>>>> Does it happen if you use javaw instead of java ?
>>> Same behavior: splash screen, then back to the prompt.
>>>> -phil.
>>>>
>>>> On 04/21/2015 11:53 AM, Pete Brunet wrote:
>>>>> Starting with b58 with JAWS (screen reader) running on Win 8, when
>>>>> starting SwingSet2 the splash screen appears but the application
>>>>> doesn't
>>>>> start.  There are no messages in the console.  I see b58 only had one
>>>>> fix:
>>>>>
>>>>> 8076531 infrastructure Switch default compiler on Windows to VS2013
>>>>>
>>>>> What might be wrong?
>>>>>
>>>>> How do I debug this?
>>>>>
>>>>> Pete
>




More information about the build-dev mailing list