RFR: 8146009: "pure virtual method called" with using new GC logging mechanism

[David Holmes]

Hi Marcus,

On 8/10/2016 12:26 AM, Marcus Larsson wrote:
> Hi,
>
> Making another attempt to fix this issue.
>
> Summary:
> The following patch resolves a problem where the VM would crash during
> shutdown due to static log related memory being de-initialized before
> the last use of the logging framework. The solution involves parts of
> the Nifty Counter idiom [0] to control static initialization and
> de-initialization of stdout's and stderr's LogOutputs. Both objects are
> now allocated using placement new, and avoids destructor calls during
> de-initialization. The LogStreamInitializer makes sure both objects are
> initialized before first use.

Exactly when will this initialization take place? You have:

! LogFileStreamInitializer::LogFileStreamInitializer() {
!   if (Atomic::cmpxchg(1, &initialized, 0) == 0) {

but on many platforms cmpxchg is handled via a stub and so the stub 
generator needs to have been initialized. If you call before that you 
will get a non-atomic implementation ... in which case no point having 
this pretend to be thread-safe.

I think I've queried this before, but does this have to rely on static 
initialization? It seems fragile and error prone.

Thanks,
David

> Because the LogOutput::Stdout/err pointers could no longer be kept in
> LogOutput, I've replaced all usages of them with the new references
> instead.
>
> The patch includes a regression test for this issue, contributed by
> Michail Chernov.
>
> Webrev:
> http://cr.openjdk.java.net/~mlarsson/8146009/webrev.00
>
> Issue:
> https://bugs.openjdk.java.net/browse/JDK-8146009
>
> Testing:
> JPRT testset hotspot, included test on supported platforms.
>
> Thanks,
> Marcus
>
> [0] https://en.wikibooks.org/wiki/More_C%2B%2B_Idioms/Nifty_Counter