RFR: 8072147: Preloading libjsig.dylib causes deadlock when signal() is called

david buck david.buck at oracle.com
Mon Jun 29 14:35:34 UTC 2015


Hi Dmitry!

Ouch. Yes, that is *very* wrong. I will fix and retest.

Sorry for the trivial error and thank you for the catch!

Cheers,
-Buck

On 2015/06/29 23:31, Dmitry Dmitriev wrote:
> Hello David,
>
> In call_os_signal function you add "reentry = false;"(line 93) after
> return statement and it seems as unreachable code.
>
> Thanks,
> Dmitry
>
> On 29.06.2015 17:18, david buck wrote:
>> Hi!
>>
>> Please review my fix below:
>>
>> bug report: https://bugs.openjdk.java.net/browse/jdk-8072147
>>
>> webrev: http://cr.openjdk.java.net/~dbuck/8072147/webrev.00/
>>
>> The fix is intentionally limited only to the platform (OSX/BSD) and
>> scenario (signal() implemented by means of a direct call to
>> sigaction()) where we can reasonably expect to see (and therefor test)
>> this issue.
>>
>> Fix has been manually tested for effectiveness by myself. "hotspot"
>> test suite run and passed on all platforms.
>>
>> Cheers,
>> -Buck
>


More information about the hotspot-runtime-dev mailing list