8257885: [TESTBUG] java/foreign/TestSegments.java fails on x86_32

[DamonFool]

Hi all,

java/foreign/TestSegments.java fails on x86_32 due to '-Xmx4G'.
The reason is that -Xmx4G is invalid maximum heap size for 32-bit platforms.
The current implimentation only supports maximum 3800M on 32-bit systems [1].

The fix just change '-Xmx4G' to '-Xmx3500M'.

Testing:

• java/foreign/TestSegments.java on Linux/x86_{32,64}

Thanks.
Best regards,
Jie

[1] https://github.com/openjdk/jdk/blob/master/src/hotspot/os/posix/os_posix.cpp#L567

---

Progress

• Change must not contain extraneous whitespace
• Commit message must refer to an issue
• Change must be properly reviewed

Issue

• JDK-8257885: [TESTBUG] java/foreign/TestSegments.java fails on x86_32

Download

$ git fetch https://git.openjdk.java.net/jdk pull/1689/head:pull/1689
$ git checkout pull/1689

[bridgekeeper]

👋 Welcome back jiefu! A progress list of the required criteria for merging this PR into master will be added to the body of your pull request. There are additional pull request commands available for use with this pull request.

[DamonFool]

/issue add JDK-8257885
/test
/label add core-libs
/cc core-libs

[openjdk]

@DamonFool This issue is referenced in the PR title - it will now be updated.

[openjdk]

@DamonFool
The core-libs label was successfully added.

[openjdk]

@DamonFool The core-libs label was already applied.

[mlbridge]

Webrevs

• 00: Full (2eccb40)

[shipilev]

-Xm3800M could still fail on many 32-bit systems, for example where native libraries are located in the middle of address space. I think it is safer to @require it. In fact, I did it an hour ago here #1688 :)

[DamonFool]

-Xm3800M could still fail on many 32-bit systems, for example where native libraries are located in the middle of address space. I think it is safer to @require it. In fact, I did it an hour ago here #1688 :)

The test just limits the maximum memory to be used, not the minimum memory.
So I think the foreign api should also work on 32-bit platforms.

[shipilev]

-Xm3800M could still fail on many 32-bit systems, for example where native libraries are located in the middle of address space. I think it is safer to @require it. In fact, I did it an hour ago here #1688 :)

The test just limits the maximum memory to be used, not the minimum memory.
So I think the foreign api should also work on 32-bit platforms.

Well, I think the test still fails with your fix: https://github.com/DamonFool/jdk/runs/1516430124 -- as I suspected it would :(

[DamonFool]

-Xm3800M could still fail on many 32-bit systems, for example where native libraries are located in the middle of address space. I think it is safer to @require it. In fact, I did it an hour ago here #1688 :)

The test just limits the maximum memory to be used, not the minimum memory.
So I think the foreign api should also work on 32-bit platforms.

Well, I think the test still fails with your fix: https://github.com/DamonFool/jdk/runs/1516430124 -- as I suspected it would :(

OK.
It seems the test is brittle on 32-bit platforms.
I agree with you now.
Thanks.