[aarch64-port-dev ] hg: aarch64-port/jdk8u/hotspot: 8157306: Random infrequent null pointer exceptions in javac

[Andrew Haley]

On 09/08/16 21:15, Andrew Hughes wrote:
> Ok, but this does result in a deviation from upstream for other
> architectures using this tree.

I get that.  I thought very long and hard about this, given the "no
divergence from upstream" rule, but in the end I decided that to
#ifdef it would be adhering to the rule to the point of ridiculousness.
It's a single very small change which isn't likely to cause
significant divergence.

If, though, it is absolutely necessary to do so in order to maintain
the rule, I'm quite prepared to #ifdef AARCH64.  However, this does
increase slightly the probability of incorrect code on other
architectures.  And, of course, this is the AArch64 port.

> I don't think this is a bad thing, but I would appreciate some
> clarification on what is appropriate for this tree. As a case in
> point, Zero is broken (crashes on -version) in this tree and will
> remain so until we pick up 8154210 from 8u112, probably in early
> November after the October CPU. Is this to be expected, or should we
> consider including such fixes which cause major runtime crashes?

Do you not mean including such fixes to major runtime crashes?

But importing critical fixes already in jdk8u is OK, I think.  I guess
we must already be patching this in the packages we ship from Red Hat.

Andrew.