Hackbench reports enormous results

[Ciprian Barbu]

The hackbench test suite reported some very peculiar results on one CI loop run, 3678167542.4300000000 seconds. This is most likely a problem with an uninitialized variable or an overflow.

http://validation.linaro.org/dashboard/streams/public/team/linaro/ci-linux-lng-preempt-rt/bundles/b45f6ef790b8b00170ad6248c6bb9186deba32f3/698b57a0-ac5b-4029-aac4-558aa50b9935/

There is a similar problem in one of the ltp tests, so solving this bug might help us finding a solution to the other one.

[Mike Holmes (mike-holmes)]

This lava filter highlights the issue http://validation.linaro.org/dashboard/image-charts/hackbench

[Mike Holmes (mike-holmes)]

Patch submitted to gerrit on staging.

[Ciprian Barbu (ciprian-barbu)]

There is a bug in hackbench, where a time variable is not initialized. I did a fix and tested it locally but I haven't upstreamed it yet. This should be done.

However, the way hackbench was being used was incorrect, so the first thing I could do was to fix that. The change was sent for review to gerrit and we are waiting for the fix to be applied:

https://staging.review.linaro.org/#/c/411/

[Mike Holmes (mike-holmes)]

Ciprian, has your fix been applied in Linaro, and what is the status of up streaming ?

[Ciprian Barbu (ciprian-barbu)]

It looks like the fix for hackbench.yaml has not been applied, I added a question in the gerrit reviewtask.

As for upstreaming the fix for hackbench itself, I will take the time to do it in the coming week.

[Ciprian Barbu (ciprian-barbu)]

I have sent a new patch set to gerrit, Anders has reviewed it with +1. Someone else will have to look at it I think.

I also sent a patch upstream for fixing the hackbench bug and the maintainer has agreed to integrate it into next release:

https://lkml.org/lkml/2013/12/12/270

[Mike Holmes (mike-holmes)]

Still in review, needs pursuing https://review.linaro.org/#/c/411/

[Ciprian Barbu (ciprian-barbu)]

Fix has been integrated, this bug should be closed.