RFR: 8252406: Introduce Thread::as_Java_thread() convenience function

David Holmes dholmes at openjdk.java.net
Mon Sep 7 21:47:48 UTC 2020


On Mon, 7 Sep 2020 13:37:10 GMT, Aleksey Shipilev <shade at openjdk.org> wrote:

>> This is a rather large but generally simple cleanup.
>> 
>> We use a lot of raw C-style casts to "(JavaThread*)" typically after checking "thread->is_Java_thread()". To simplify
>> this pattern, and make the code look somewhat cleaner we introduce a convenience function Thread::as_Java_thread(),
>> which asserts is_Java_thread() and returns "this" cast to "JavaThread*". A const version, as_const_Java_thread(), was
>> also added to allow use in cases where we start with "const Thread *".  Summary of changes:
>> - changed raw casts to use as_Java_thread()
>> - removed redundant checks of is_Java_thread() as it is now done in as_Java_thread()
>> - Removed checks that are checking the type system e.g.
>> void foo(JavaThread* t) {
>>   assert(t->is_Java_thread(), "must be")
>> the only way the assert can fire is if someone deliberately bypasses the type system, and if we are going to worry
>> about that then we would need asserts like this on every method that expects a JavaThread! The right place for the
>> assertion is at the head of any such call chain.
>> - Removed asserts that are already guaranteed in the caller.
>> - Use JavaThread::current() where appropriate.
>> - Use pre-existing "thread" variable where available rather than casting THREAD
>> 
>> Change locations found by grepping for variations of the cast syntax "(JavaThread*)" - it is possible some may have
>> been missed.
>> Testing: tiers 1-3
>> 
>> Thanks,
>> David
>
> Drive-by comment: Shenandoah changes look good.

Thanks @shipilev ! Sending this out for RFR now.

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

PR: https://git.openjdk.java.net/jdk/pull/37


More information about the serviceability-dev mailing list