Skip to main content

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index] [List Home]
Re: [jdt-core-dev] timeout

Where can we see / change the -Xmx used during tests?

*If* this is a memory issue, is it a good idea to ignore the
signal from tests and just accept the higher consumption?

best,
Stephan


On 02/27/2017 09:06 PM, Andrey Loskutov wrote:
Hi Stephan,

I imagine that max heap default to -Xmx1024m used for SDK is too less for the old + new index, and that the GC has hard times to
free the memory like in bug 512749 (java.lang.OutOfMemoryError: GC overhead limit exceeded).

I had already reported this via bug 511693, that for my personal setup the memory consumption seem to be doubled in 4.7. I don't see
OOM errors but only because I have max heap increased to 3G. In my typical setup, in the idling Eclipse with ~10 opened editors and
lot of Platform projects in the workspace, JDT uses 520 from 804 MB (used by the JVM), and more then a half of that (270 MB) is used
by the org.eclipse.jdt.internal.core.nd.db.Chunk objects.

But of course it can be that jdt Hudson has some issues. Have you tried to restart it (in case Hudson leaks some memory or file
descriptors or both)?

Am 27.02.2017 um 20:44 schrieb Stephan Herrmann:
Hi folks,

Gerrit is of very little help these days, timing out more often then not.

Of the last 15 runs, only one succeeded (#2462), which contained this:

----
combined with:
Bug 512740: [newindex] preference to disable the new index does not
disable its rescan job

AND: HARD-DISABLE the new index for now
----

Is that just a coincidence or is the new index really causing our tests
to time out?

(Locally on my machine tests almost always hang during AllJavaModelTests
if the new index is enabled).


Or has s.t. else in the stack degraded in performance?

Some random references to candidate culprits:

HIPP:
https://dev.eclipse.org/mhonarc/lists/cross-project-issues-dev/msg14119.html


OOME: https://bugs.eclipse.org/bugs/show_bug.cgi?id=512749

Seeing the compiler in the stack reminds me also of:
https://bugs.eclipse.org/bugs/show_bug.cgi?id=512156


any theories?
Stephan




Back to the top