www-builds mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Jesse Glick <typr...@gmail.com>
Subject Re: Apache build infrastructure or Oracle JVM problem: crash in native JDK code
Date Wed, 20 Feb 2013 17:12:52 GMT
On 02/20/2013 11:46 AM, Martin Desruisseaux wrote:
> clobbering an open JAR file is an hypothesis that may explain the JVM crash, but this
is unverified.

Indeed it is only a hypothesis, though one based on repeated encounters with similar crashes
over the course of years of development that were generally traced back to 
this cause.

> nothing is touching this JAR file in parallel

Right, there is no known parallel build. There are other less likely explanations, such as
asynchronous writes happening in response to garbage collection, etc. A real 
investigation would require tracking OS writes or something like that.

I would merely claim that loading classes from a snapshot JAR is _prone_ to triggering crashes
when other factors come into play which might be difficult to predict. And 
it is certainly not out of the question that different JDK versions behave in subtly different
ways that are enough to make your crash happen or not; this does not mean 
that the bug is only present in 1.7.0.

> This snapshot JAR file is used as a plugin for other modules well after the build has
been completed.

Well after the build of that plugin module has been completed I suppose you mean. (According
to your build log, the JAR is created earlier in the reactor build and then 
loaded at the moment of the crash.)

At any rate, for many reasons it would be preferable to run on the latest available version
of Java (7u15 as of this writing). I do not personally have admin permissions 
to help you there.

Mime
View raw message