Review request for 5049299

Andrew Haley aph at redhat.com
Fri May 29 10:54:37 UTC 2009


Florian Weimer wrote:
> * Andrew Haley:
> 
>> Florian Weimer wrote:
>>> * Andrew Haley:
>>>
>>>>> The latter leads to problems if the maximum heap size is relatively large.
>>>>> Currently, -Xmx is reserved, and some software fails to build on
>>>>> vm.overcommit_memory systems due to this.
>>>> I don't get it.  Why would overcommit cause some software to fail to
>>>> build?  I can understand it failing if overcommit was turned off.
>>> Sorry, I meant to write "vm.overcommit_memory=2".
>> Well, the system is doing what the user asked it to do.  -Xmx means
>> "please reserve this memory for me, Hal"; if the memory isn't there,
>> then it's surely quite reasonable for Hal to respond "I'm sorry Dave,
>> I can't do that."
> 
> Well, I thought -Xms means that.  An untuned VM is not expected to hit
> the -Xmx limit.  At least this is what I see in practice, and to me
> this is the distinction expressed between the -Xmx and -Xms flags.

I am very sorry, "-Xmx" was a typo, or perhaps a thinko.  In my original
posting -- which was trimmed -- I said

> It makes more sense to allocate all the -Xms size immediately.

and I thought you were disagreeing with me, but:

> Reserving the -Xms heap portion does make sense, though.

you weren't.

Andrew.




More information about the core-libs-dev mailing list