RFR: 8301306: java/net/httpclient/* fail with -Xcomp: java.io.IOException: HTTP/1.1 header parser received no bytes

[Vladimir Kozlov]

On Mon, 30 Jan 2023 21:13:12 GMT, Daniel D. Daugherty <dcubed at openjdk.org> wrote:

>> Hi all,
>> I think `BackgroundCompilation` should not be set to false when `-Xcomp` be used, which causes the java thread to block for longer, then causing the following tests failed frequently on the AArch64 and LoongArch64 architecture.
>> 
>> java/net/httpclient/ByteArrayPublishers.java （fastdebug -Xcomp）
>> java/net/httpclient/ManyRequestsLegacy.java （fastdebug -Xcomp）
>> java/net/httpclient/HttpClientLocalAddrTest.java （fastdebug -Xcomp）
>> jdk/incubator/concurrent/StructuredTaskScope/StructuredTaskScopeTest.java （fastdebug -Xcomp）
>> java/rmi/server/UnicastRemoteObject/serialFilter/FilterUROTest.java （release -Xcomp）
>> 
>> This PR delete BackgroundCompilation be setted, Above tests can be passed. Please help review it.
>> 
>> Thanks.
>
> src/hotspot/share/runtime/arguments.cpp line 1440:
> 
>> 1438:   case _comp:
>> 1439:     UseInterpreter           = false;
>> 1440:     BackgroundCompilation    = false;
> 
> Disabling `BackgroundCompilation` when -Xcomp is specified is not the right
> answer here. When `-Xcomp` is specified, `-Xbatch` needs to be turned off
> and I believe that's what old L1440 is doing.

Dan is right, you  can't change the definition of -Xcomp.

But `-Xbatch` is `-XX:-BackgroundCompilation`. So the second Dan's statement is incorrect.

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

PR: https://git.openjdk.org/jdk/pull/12282