RFR: 8011661: Insufficient memory message says "malloc" when sometimes it should say "mmap"
coleen.phillimore at oracle.com
Tue Apr 30 07:17:59 PDT 2013
This code looks good to me. I thought the same thing as David but the
distribution wasn't all malloc and a few mmaps, or the other way
around. The change is disruptive, but I think it's helpful. Our long
term goal is to reduce the number of these calls in favor of something
On 04/30/2013 02:36 AM, David Holmes wrote:
> Hi Calvin,
> This would be far less disruptive if you used a default parameter so
> that the majority of the call sites did not need to be changed. Also
> the use of an enum seems excessive unless you think we will discover
> other OOM reasons ?
> On 30/04/2013 11:37 AM, Calvin Cheung wrote:
>> Thanks for your review.
>> Since the fix for 8011773 (Some tests on Interned String crashed JVM
>> with OOM) has been pushed today, I needed to update the fix to include
>> the change in oop.cpp. Webrev is at the same location:
>> Also, I need one more reviewer so that this fix can be pushed.
>> I ran the vm.quick.testlist on linux_x64, there's no new failures
>> compared with the promoted build.
>> On 4/29/2013 7:01 AM, Zhengyu Gu wrote:
>>> Good to me.
>>> On Apr 26, 2013, at 6:32 PM, Calvin Cheung wrote:
>>>> Problem: needs a way to differentiate the 2 OOM error cases - malloc
>>>> or mmap
>>>> Adding an enum parameter to the vm_exit_out_of_memory macro and the
>>>> function so that the caller can indicate whether it's a malloc or
>>>> mmap error.
>>>> Main changes are in debug.[cpp | hpp] and vmError.[cpp | hpp].
>>>> Other changes are mainly for passing in the additional enum when
>>>> calling vm_exit_out_of_memory.
>>>> Test coverage:
>>>> webrev: http://cr.openjdk.java.net/~ccheung/8011661/webrev/
>>>> jbs: https://jbs.oracle.com/bugs/browse/JDK-8011661
More information about the hotspot-dev