[aarch64-port-dev ] JTreg failures in aarch64 jdk7u

[Andrew Haley]

Hi,

On 07/24/2014 11:01 AM, Edward Nevill wrote:

> I have been looking at the hotspot jtreg failures in aarch64 jdk7u. Currently I get.

We are already having a conversation about this.  See
"jdk8 test cases in the jdk7u archives" in this list.

> Test results: passed: 343; failed: 102; error: 1
> 
> This is because the hotspot in aarch64 jdk7u is based on jdk8
> hotspot (hs25) instead of the jdk7u hotspot (hs24). The jtreg test
> suite in aarch64 jdk7u contains many tests which fail because they
> use JDK8 source features, or depend on JDK8 APIs.
> 
> For example, all 38 mathexact tests fail because there is no
> mathexact in JDK7.
> 
> Running the same test suite on x86 gives me
> 
> Test results: passed: 311; failed: 134; error: 1
> 
> If I take the jtreg hotspot test suite from x86 jdk7u hotspot (hs24)
> and run it on aarch64 jdk7u I get the following results.
> 
> Test results: passed: 279; failed: 11; error: 2
> 
> compared with the following running it on x86
> 
> Test results: passed: 286; failed: 4; error: 2
> 
> So, what is the best solution to this?
> 
> 1) Prune the existing aarch64 jdk7u hotspot jtreg tree to remove all
> tests which depend on jdk8 or
> 2) Generate an exclude list of the above tests or
> 3) Replace the aarch64 jdk7u hotspot jtreg tree (from hs25) with the
> one from x86 (hs24).

I think that 3) would be best.

Medium term, we should do a real backport of AArch64 to hs24 and
integrate it with the current icedtea7-forest tree.  Then,
icedtea7-forest users get to build from a single set of sources.

Andrew.