Question about JDK-8007993 - TraceDependencies with enabled LogCompilation makes hotspot.log invalid
Pavel Punegov
pavel.punegov at oracle.com
Thu Mar 7 09:10:34 PST 2013
Hi Vladimir,
could you please or someone take a look at this [*] and evaluate the problem?
We have some compiler tests which depend on hotspot.log file, and they could
fail the same way as in the bug.
[*] https://jbs.oracle.com/bugs/browse/JDK-8007993
Thanks,
Pavel
On Thursday 07 March 2013 21:02:57 Pavel Punegov wrote:
> The same caused failures described in this bug. There are no synchronization
> in xmlStream and hence when multiple threads are writing to TTY hotspot.log
> became invalid.
>
> In JDK-8007993 nmethod and Dependencies classes were writing concurrently
> from different threads.
>
> So I think these are the same issues. Please add all information you have
> and maybe we should rename bug to smth like "hotspot.log writing
> synchronization issues".
>
> On Thursday 07 March 2013 20:46:12 Filipp Zhinkin wrote:
> > On 03/07/2013 07:33 PM, Pavel Punegov wrote:
> > > I don't remember if I tried to run it with TraceDependencies off, but
> > > issue described in bug was caused by dependency output from this option.
> > >
> > > How does it reproduce on your desktop and what is hotspot.log generated?
> >
> > Tests failing due to invalid xml hotspot.log with fastdebug hs20, hs24,
> > hs25 with -Xcomp mode only.
> > At the first glance issues could be caused by the lack of
> > synchronization in logging mechanism, because hotspot.log contains some
> > corrupted elements, like multiple threads were writing into it
> > simultaneously.
> >
> > > On Thursday 07 March 2013 19:10:58 Filipp Zhinkin wrote:
> > > > Pavel,
> > > >
> > > >
> > > >
> > > > why do you think that this issue caused by TraceDependencies?
> > > >
> > > > On my desktop issues could be reproduced even with TraceDependencies
> > > >
> > > > explicitly turned off. Or it is yet another issue? :)
> > > >
> > > >
> > > >
> > > > Thanks,
> > > >
> > > > Filipp.
> > >
> > > Thanks,
> > >
> > > Pavel
-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://mail.openjdk.java.net/pipermail/hotspot-compiler-dev/attachments/20130307/fe965b82/attachment.html
More information about the hotspot-compiler-dev
mailing list