Re: How to do the sunrpc revert?

["Carlos O'Donell" <carlos at systemhalted dot org>]

On Wed, May 9, 2012 at 5:37 AM, Allan McRae <allan@archlinux.org> wrote:
> On 09/05/12 18:17, Andreas Jaeger wrote:
>> I looked briefly at the sunrpc changes while waiting for more people to give
>> their "Yes, do it" ;).
>>
>> Fedora and openSUSE (not checked others) use a very simple patch (attached as
>> glibc2.14-revert-sunrpc-removal.patch) that works but is more of a hack, e.g.
>> it does the following in include/libc-symbols.h:
>>
>> -# define libc_hidden_nolink(name, version) hidden_nolink (name, libc, version)
>> +# define libc_hidden_nolink(name, version) hidden_def (name)
>>
>> The full removal patch (attached as sunrpc-removal.patch) is rather large and
>> contains also whitespace changes which I propose not to revert.
>>
>> Should we do a full revert of the patch or use the simple patch - or is there
>> some middle ground?
>>
>> If we do a revert, what should not be reverted besides copyright lines
>> and whitespace changes?
>>
>> I consider the simple patch a bit too hackish.
>
> It might be hackish, but it has the advantage of being quite well tested
> (I'll add Arch and Gentoo to the list of distros that use the same
> patch) which with less than a week until 2.16 freeze would be good.
> Also, it is very small so will be easy to revert when this really can be
> removed.

As the 2.16 release manager I would agree with Allan here.

* Apply hack to trunk.

* Release 2.16 with well tested hack.

* File a bug in with milestone=2.17.

* Revert it fully after 2.16 branches.

I know that makes your life a little more complicated but it's
certainly less risky.

Cheers,
Carlos.