[2.22] Backport malloc fixes from BZ#19048 and BZ#19243

[Paul E. Murphy]

On 04/13/2016 01:15 PM, Carlos O'Donell wrote:
> On 04/13/2016 01:48 PM, Florian Weimer wrote:
>> On 04/13/2016 06:35 PM, Paul E. Murphy wrote:
>>>
>>>
>>> On 04/13/2016 06:49 AM, Florian Weimer wrote:
>>>> On 04/12/2016 11:53 PM, Paul E. Murphy wrote:
>>>>> This attempts to backport BZ#19048, and the related fix in BZ#19243.
>>>>>
>>>>> The TLS refactoring made backporting 19048 a little more than
>>>>> trivial.  It did not raise errors after running make check for x86_64
>>>>> and ppc64le.
>>>>>
>>>>> This should get a review before pushing.
>>>>
>>>> NEWS has unrelated changes.
>>>>
>>>> You should also backport commits 90c400bd4904b0240a148f0b357a5cbc36179239, 7962541a32eff5597bc4207e781cfac8d1bb0d87, and perhaps 2a38688932243b5b16fb12d84c7ac1138ce50363.
>>>
>>> Thanks for pointing those out.  I've backported the above 3 as well, and fixed
>>> NEWS (oops).
>>
>> I compared arena.c, malloc.c, and Makefile, and the differences look okay to me.
>>
>>> Is there any potential objection to pushing these onto the common 2.22 release
>>> branch? Otherwise, I'll keep them on the ibm branch.
>>
>> We've backported the same set of changes fairly widely, too. I expect
>> it to be a low-risk change at this point. Fedora currently doesn't
>> use the 2.22 branch, though, so I do not feel qualified to approve
>> this.
> 
> No approval is required.
> 
> If the changes are on master they are automatically approved for
> stable branches *unless* someone objects on the grounds that they
> are too unstable.
> 
> Given that the patches fix internal malloc algorithm issues I don't
> see any reason why we should not backport them aggresively.
> 

Thanks all for the help. These are now committed to the 2.22 release branch.