RFR: 8024671: G1 generates assert error messages in product builds

Aleksey Shipilev aleksey.shipilev at oracle.com
Thu Sep 12 12:37:21 UTC 2013


On 09/12/2013 04:30 PM, Thomas Schatzl wrote:
>  thanks for finding and fixing this. Also for catching the typo.

Thanks for the review, Thomas!

> On Thu, 2013-09-12 at 15:59 +0400, Aleksey Shipilev wrote:
>> Hi,
>>
>> Please review and push this simple change:
>>   http://cr.openjdk.java.net/~shade/8024671/webrev.00/
>>
>> TL;DR: err_msg() is not getting erased in product builds, because
>> check_card_num is not a macro, causing the performance regression. Turns
>> out it is simpler to inline the check_card_num all along, because it is
>> more consistent to other asserts.
> 
> I really do not like the inlining of the check_card_num() method,
> replicating the code. I would prefer an approach that avoids that.

I thought that too before starting the patch, targeting to declare the
proper macro for check_card_num. But, inline asserts seem be coherent
will other asserts in the affected code blocks. Hence, I changed my
mind, and produced the current patch.

-Aleksey.




More information about the hotspot-gc-dev mailing list