Backport vmTestbase tests to jdk8u-dev?

[Liu, Xin]

Hi, Maintainers, 

Thank you to review the patches. I put answers inline. 

On 12/29/19, 8:01 PM, "Andrew John Hughes" <gnu.andrew at redhat.com> wrote:
    
    On 29/12/2019 19:35, Liu, Xin wrote:
    > Hi, Maintainers, 
    > 
    > I put answers inline.
    > 
    > Please provide the following info for this work:
    >    1. Prepare patches for all affected repositories. List the repositories
    >       affected. Send a link to patches which apply to current jdk8u-dev
    >       HEAD.
    > we will touch two repos. 
    > a) jdk8u-dev repo: 
    >     1. https://cr.openjdk.java.net/~xliu/8231089/root/webrev/
    > b) jdk8u-dev/hotspot
    >     1. base                         https://cr.openjdk.java.net/~xliu/8231089/base.patch.tar.gz
    >     2. updateAPIs             https://cr.openjdk.java.net/~xliu/8231089/updateAPIs.patch
    >     3. classFileInstaller    https://cr.openjdk.java.net/~xliu/8231089/classFileInstaller/webrev/
    >     4. gclog                        https://cr.openjdk.java.net/~xliu/8231089/gclog/webrev/
    >     5. nativeLibraries       https://cr.openjdk.java.net/~xliu/8231089/nativeLibraries/webrev/
    >     6. azul_contrib1         https://cr.openjdk.java.net/~xliu/8231089/azul_contrib.patch
    >     7. azul_contrib2         https://cr.openjdk.java.net/~xliu/8231089/azul_contrib2.patch
    >     8. new_targets           https://cr.openjdk.java.net/~xliu/8231089/new_targets/01/webrev/
    > 
    
    Where does all this originate from? JDK-8231089 seems to be a freshly
    created bug rather than an issue resolved in later JDK versions being
    backported.
    
I start with these issues.
https://bugs.openjdk.java.net/browse/JDK-8202812?jql=labels%20%3D%20test-opensource

but they are so scattered,  I took a different approach. I did a blob copy of vmTestbase directory from tag jdk-14+14. It's the patch #1. 
And then gradually make vmTestbase work.  All others are not backports.  eg. patch #2 tailors for JDK8 APIs.
Patch #4 modifies Java arguments for gclog because jdk8 doesn't have unified logging. 

We also don't bring the new building system to jdk8, but vmTestbase does need to compile native auxiliary libraries. We support it by patch #5. 
We don't support 'make test' at all. To work around it, we extend the makefile of hotspot to trigger vmtestbase. Please refer to the patch #8. 

FYI, I published the basic layout here.  
https://mail.openjdk.java.net/pipermail/jdk8u-dev/2019-August/010173.html
even though many details have changed, the basic idea is still the same.  

    >    2. Tell us on which platforms and architectures you've built/run those
    >       tests.
    > I ran on linux_x86-64 and linux_aarch64.
    > 
    >    3. Send a diff stat of the patches. This is to ensure only test
    >       directories are being touched.
    > Please refer to the attachment vmTestbase.status.zip. it enlists all changed files. 
    > I understand your concern. That's our intention too. I prefer to keep the file layout intact.  It's because it would be easier to update vmTestbase in the future. 
    > 
    >    4. Provide instructions as to how these tests can be run.
    > Please refer to the script attached. Two functions are provided. 
    > apply_vmTestbase apply patches. 
    > run_vmTestbase kicks  off vmTestbase test.
    > 
    >    5. The patch should eventually get pushed under a single bug, JDK-
    >       8231089. That means adding vmTestbase code should be a single big
    >       commit when pushed. So please squash any changes you may have.
    > Sure. I will do that. The reason I'd like to keep mq format now because it's easy to modify commits.  
    > 
    >   6. The commit message should perhaps also mention "JDK-8199257:
    >       [TESTBUG] Open source VM testbase metaspace tests" as it also
    >       includes those? An eventual push would then list a backport of this
    >       bug.
    > Yes, it includes.  I copy vmTestbase blob from jdk-14+14. I also double-check the log. It includes those metaspace changes. 
    > 
    > For test stats, so far, it passed 3480 tests and I put 82 tests in a ProblemList(https://cr.openjdk.java.net/~xliu/8231089/new_targets/01/webrev/test/ProblemList-vmTestbase.txt.html)
    > 
    
    Why is JDK-8199257 not the patch being backported here?
    
    Thanks,

JDK-8199257 is included.  The attachment is my running log of vmTestbase. It includes vmTestbase/metaspace/*.
The reason they disable some gc/firstGC teste because of https://bugs.openjdk.java.net/browse/JDK-8208250. 
But actually, they work fine on jdk8u. 

Thanks,
--lx

    -- 
    Andrew :)
    
    Senior Free Java Software Engineer
    Red Hat, Inc. (http://www.redhat.com)
    
    PGP Key: ed25519/0xCFDA0F9B35964222 (hkp://keys.gnupg.net)
    Fingerprint = 5132 579D D154 0ED2 3E04  C5A0 CFDA 0F9B 3596 4222
    https://keybase.io/gnu_andrew