8226863: [aix] jdk/java/lang/reflect/exeCallerAccessTest cannot launch on primordial thread

Thomas Stüfe thomas.stuefe at gmail.com
Fri Jun 28 13:36:14 UTC 2019


Hi Christoph,

On Fri, Jun 28, 2019, 13:19 Langer, Christoph <christoph.langer at sap.com>
wrote:

> Hi Thomas,
>
>
>
> I looked at this.
>
>
>
> One thing that catches my eye immediately: exeCallerAccessTest.c, line
> 149: Looks like this #error would always fire on Windows. So the test
> wouldn’t compile there…? Or am I wrong?
>
>
>

No :(

That whole code section was supposed to be conditional on
DEFAULT_SPAWN_IN_NEW_THREAD.

Must be the heat...

Furthermore, if we do this new thread dance only on AIX, shouldn’t all the
> extra coding be guarded by #ifdef _*AIX? *The only common part of the
> changes would be to rename main to main_inner and call this method for all
> non-AIX cases…
>

Yea, I'll repost an update next week. Thanks for the review.

Cheers, Thomas


>
> Best regards
>
> Christoph
>
>
>
> *From:* ppc-aix-port-dev <ppc-aix-port-dev-bounces at openjdk.java.net> *On
> Behalf Of *Thomas Stüfe
> *Sent:* Donnerstag, 27. Juni 2019 09:02
> *To:* Java Core Libs <core-libs-dev at openjdk.java.net>; ppc-aix-port-dev <
> ppc-aix-port-dev at openjdk.java.net>
> *Subject:* 8226863: [aix] jdk/java/lang/reflect/exeCallerAccessTest
> cannot launch on primordial thread
>
>
>
> Hi all,
>
>
>
> Issue:
>
> https://bugs.openjdk.java.net/browse/JDK-8226863
>
> webrev:
>
>
> http://cr.openjdk.java.net/~stuefe/webrevs/8226863--jdk-java-lang-reflect-execalleraccesstest-cannot-launch-on-primordial-thread/webrev.00/webrev/
>
>
>
> we have this annoying issue on AIX that the libjvm cannot be invoked on a
> primordial thread. Therefore we need to change launchers which create the
> VM to spawn own threads (we only do this where it is worth the effort -
> this, admittedly, is a corner case, we could alternatively just disable the
> test on AIX).
>
>
>
> This is annoying and vaguely embarrassing :-/ Maybe IBM will step in some
> time and help us solve the underlying issues which have to do with the
> inability to create guard pages on the primordial thread stack (among other
> things IIRC).
>
>
>
> Thanks for reviewing,
>
>
>
> Thomas
>
>
>
>
>


More information about the core-libs-dev mailing list