[jdk11u-dev] RFR: 8237354: Add option to jcmd to write a gzipped heap dump

[Paul Hohensee]

On Mon, 9 Aug 2021 07:08:36 GMT, BuddyLiao <github.com+3094961+buddyliao at openjdk.org> wrote:

> Backport of the heap dump enhancement: Add option to jcmd to write a gzipped heap dump
> 
> Patch did not apply cleanly, I had to resolve/modify
> 
> 1) src/hotspot/share/services/heapDumper.cpp
> function "int dump(const char* path, bool overwrite = false)" has some different from the jdk-master, since patch 7cbb67a3f8adc83a5b51c092a66480d7b22a6bea is backport before this one, so this patch can't backport clearly
> 
> 2) src/hotspot/share/services/heapDumperCompression.cpp
> MonitorLockerEx is named MonitorLocker in jdk-master, MonitorLockerEx.wait should have parameters, the constructor of  MonitorLockerEx and MutexUnlockerEx also different from jdk-master
> 
> 3) src/hotspot/share/services/heapDumperCompression.hpp
> FileWriter is different from jdk-master since backport of 7cbb67a3f8adc83a5b51c092a66480d7b22a6bea before this patch.
> 
> I have built it on linux x86_64 and run test-tier1

I applied the patch and ran the serviceability/dcmd/gc tests with "make run-test". HeapDumpAllTest crashed due to the lack of the the line 258 change. Adding that cured that crash, but the new HeapDumpCompressedTest test fails due to lack of a backport of [JDK-](https://bugs.openjdk.java.net/browse/JDK-8233790). 8233790 guarantees that the "Unable to create" string for which the test checks will be in the output file. That backport looks useful to me, so I'd do it as a pre-req to this one.

-------------

Changes requested by phh (Reviewer).

PR: https://git.openjdk.java.net/jdk11u-dev/pull/209