Re: [PR18457] Don't require rtld lock to compute DTV addr for static TLS

[Alexandre Oliva <aoliva at redhat dot com>]

On Jun  4, 2015, Torvald Riegel <triegel@redhat.com> wrote:

> On Wed, 2015-06-03 at 17:31 -0300, Alexandre Oliva wrote:
>> On Jun  3, 2015, Alexandre Oliva <aoliva@redhat.com> wrote:
>> 
>> > How's this?
>> 
>> Or rather this, that I got after updating the patch file :-)

> Why did you remove the atomic accesses altogether?

Because I rearranged the code so that the double-checked lock pattern is
self-evident.  I thought we had agreed long ago that we didn't need
atomics for double-checked locks.  You added an "*Correct*" in this
thread, so now I guess you have to show why the proposed change is not
correct.  Please explain?

> You made good progress towards a consistent fix with the reasoning on
> synchronization you provided for the static case

It was not for the static case only.  It covered both cases.

> This doesn't add any of the documentation I want to see either.

The deal I suggested was that I'd answer your questions and you'd write
the documentation.  Now you're moving the goalpost.  Anyway, my manager
told me not to spend time on this Q&A project I had suggested, so you'll
have to find some other way to get the documentation you want.
Stonewalling a trivial patch on the grounds that it fails to document
something that was not documented to begin with it not reasonable; it
won't even get you the documentation, it will just cause the regression
to be around longer.

-- 
Alexandre Oliva, freedom fighter    http://FSFLA.org/~lxoliva/
You must be the change you wish to see in the world. -- Gandhi
Be Free! -- http://FSFLA.org/   FSF Latin America board member
Free Software Evangelist|Red Hat Brasil GNU Toolchain Engineer