harmony-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "Stepan Mishura" <stepan.mish...@gmail.com>
Subject Re: [general][M3] release candidate status
Date Tue, 02 Oct 2007 07:24:31 GMT
On 10/2/07, Alexei Fedotov <alexei.fedotov@gmail.com> wrote:
> Hello, Stepan,
> > 3 failures on Windows and 11 on Linux - I haven't seen analysis report
> I have investigated log files for all 14 failures, and all of them are
> due to test errors. Most of them are due to timeouts, and others are
> due to OutOfMemoryError. All of these parameters are hard-coded and
> not automatically adjusted during execution, so changes in VM
> performance produce such unexpected problems.

Yes, the suite should be fixed - tests shouldn't have any hard coded timeouts.

> Even if some timeouts could be caused by Harmony implementation
> deadlocks, the correct hing is to address the tests first since they
> are much easier to fix.
> BTW,
> There are also massive failures on Win64 platform because TI is called
> with JIT. If you have a good example of ant build.xml which checks for
> 64 bit platform (or even uses platfrom-dependent excludes), this would
> be a great help.

Do you mean that the suite should be run in interpreter mode on 64 bits?
If you need an example - jdktools-test suite is run in interpreter
mode only, see its adaptor.xml.


> BTW2,
> Nearly all memory fragmentation tests created for H-4114 produce a
> crash in the function gc_heap_slot_write_ref(). Two old tests crash in
> trace_object() function. I will re-run the tests using debug build.
> Strangely, these crashes do not apear in CC report.
> Thanks.

View raw message