We have regression in inaccuracy tests from 17-10-2008.
- Launch IDE with fresh user directory
- Open Perl project
- Open config.h file in editor
==> This code is highlighted as error:
#if 42 != 1 && 42 != 42
# include "Bletch: How does this C preprocessor concatenate tokens?"
P2 because I see this problem in RC1 also.
most probably problem is in JVM. Problem is reproducible only on
java version "1.6.0_10"
Java(TM) SE Runtime Environment (build 1.6.0_10-b33)
Java HotSpot(TM) 64-Bit Server VM (build 11.0-b15, mixed mode)
everything is fine with 1.6.0_05 and 1.5.0_16
Looks like the problem with inlining of simple functions.
Reassigning to QA for evaluation of possible solutions.
Submitted P1 bug on jvm:
not a stopper for NB6.5 release as non-reproducible with previous java versions (earlier then 1.6.0_10) but should be
fixed in next release and in patch for NB6.5
OpenSolaris 11/08 is going to have jdk1.6 update 10 as the default.
Does it make this bug a showstopper?
It does not look as a showstopper under any circumstances. And in this case the issue has been reported as purely Linux
issue (may be even only Linux64 issue), so I can not believe it affects OpenSolaris.
QA may want to double check it.
public visible java bug:
just a note to QA: We can not fix this problem on our side, so not clear which fix could be included in 6.5 patch?
I have found some tricks how to change code to be not overkilled by java machine 6 Update 10
Created attachment 72928 [details]
fixup to workaround jdk bug
another workaround is added due to bug in jdk6 Update 10
Integrated into 'main-golden', will be available in build *200811011401* on http://bits.netbeans.org/dev/nightly/ (upload may still be in progress)
User: Vladimir Voskresensky <email@example.com>
Log: workaround (due to bug in jdk6u10) for IZ#150693: Regression in inaccuracy tests from 17-10-2008
Verified in 200811200201 build
The fix was ported into release65_fixes repository.
verified in nb6.5 with patch1