RFR: 8230228: [TESTBUG] Several runtime/ErrorHandling tests may fail on some platforms

[Thomas Stüfe]

Thanks for fixing, looks reasonable.

About the specific address, yes, we are aware of that. We also could
dynamically generate an invalid address by mmapping and mprotecting, but
that address would have to be communicated back to the test and that would
have made the test slightly more complicated than necessary. Basically, it
works well enough for now, and if we find that it doesn't work on all
platforms anymore we can change it.

Cheers, Thomas


On Tue, Aug 27, 2019, 14:56 Jie Fu <fujie at loongson.cn> wrote:

> Hi all,
>
> JBS:    https://bugs.openjdk.java.net/browse/JDK-8230228
> Webrev: http://cr.openjdk.java.net/~jiefu/8230228/webrev.00/
>
> These tests failed on our mips Fedora-21 platform.
> ----------------------------------------------
>
> jtreg:test/hotspot/jtreg/runtime/ErrorHandling/TimeoutInErrorHandlingTest.java
> jtreg:test/hotspot/jtreg/runtime/ErrorHandling/ErrorFileRedirectTest.java
>
> jtreg:test/hotspot/jtreg/runtime/ErrorHandling/BadNativeStackInErrorHandlingTest.java
>
> jtreg:test/hotspot/jtreg/runtime/ErrorHandling/SafeFetchInErrorHandlingTest.java
> ----------------------------------------------
>
> The reason is that memory access to address 0xABC0000000000ABCULL, which
> is returned by VMError::get_segfault_address() [1], trigers SIGBUS, not
> SIGSEGV, on our platform.
> Please note that it seems there is no way to guarantee to generate
> SIGSEGV by accessing an specific invalid memory address for all
> platforms [2].
> Could you please review it and give me some advice?
>
> Thanks a lot.
> Best regards,
> Jie
>
> [1]
>
> http://hg.openjdk.java.net/jdk/jdk/file/cb836bd08d58/src/hotspot/share/utilities/vmError.cpp#l68
> [2]
>
> http://hg.openjdk.java.net/jdk/jdk/file/cb836bd08d58/src/hotspot/os_cpu/bsd_x86/os_bsd_x86.cpp#l579
>
>
>