Re: commit reversion without review

[Alexandre Oliva <aoliva at redhat dot com>]

On May  9, 2018, "Dmitry V. Levin" <ldv@altlinux.org> wrote:

> On Tue, May 08, 2018 at 01:22:32PM -0300, Alexandre Oliva wrote:
>> On May  8, 2018, "Dmitry V. Levin" <ldv@altlinux.org> wrote:
>> 
>> > I suppose he just forgot how things are done here.
>> 
>> It is absolutely correct to state I had no idea that reverting a patch
>> that had gone in without consensus and that was still under discussion
>> required the reversal patch to be posted in any special way.  I don't
>> see any such rule spelled out.  Still, I followed the written procedure:
>> I posted the proposal, met the consensus criterion and, after several
>> days, I installed the proposed reversal.
>> 
>> I do not see how it was a mistake. [...]

> This is obvious.  What we usually commit without review is trivial bug-fix
> changes (there is a tentative list of that kind of changes in the wiki).  

Yeah, I'm familiar with that list.

> For a revert commit to fall into this category, it has to be trivial.

It was trivial indeed, but I did not rely put it in as a trivial
bug-fix, but rather as a consensual restore of initial conditions.

> Let's have a look at your commit message from this perspective.

> The first issue with this commit is technical: if you were using "git revert",

I did use git revert.  Have you ever used that in GNU libc?  There's
this thing, the ChangeLog, that can't just be reverted, and that "git
revert" will nearly always report conflicts on.

So I had to manually fix the ChangeLog.  Then, as in recommended
practice, I put that ChangeLog in the commit message, along with the
rationale for the change.

> If you'd posted your proposed commit for review

Indeed, I posted just an offer of reversal (the patch for a revert is
the same as the initial patch, applied with -R).  Others had signaled
the reversal as a possibility before ("I won't revert it myself, but if
you can find someone else to do it", quoted form memory).  I waited for
days after the offer, and no one responded to it other than Richard.  So
I understood there was consensus to restore the initial conditions and
installed it.

> The second issue is the lengthy text full of controversial statements:

You and others may have found them controversial, but it was my
rationale for the (to me) consensual and obvious fix.

Any obvious fix could carry its own rationale, and nobody would object
to that the way you did.  I wonder why.

> every paragraph starting with the second one contains at least one
> statement that was a subject of heated discussions in this list.

Every one of the arguments has an analogous argument that applies to the
initial commit, but somehow the initial commit is not controversial.  I
wonder why.

> So let me reiterate: next time please post your proposed commits for review.

I did, but AFAICT nobody read it, so nobody objected, so it went in.
What I did NOT do was to post the patch, which I still think is not
necessary for a revert, and I did NOT post it as a separate thread,
which I now regret, for it would have made it more visible, so there
wouldn't be much doubt about whether objections applied to it (they'd be
raised in response to the proposal, I suppose)

In case you missed the proposed commit, see
https://sourceware.org/ml/libc-alpha/2018-05/msg00053.html
and look at the very end:

  [...]  I offer to reverse the patch myself, if the person who pushed
  it in doesn't do so in the next few days, so that we can then seek
  consensus without the fait accompli artificially shifting the
  baseline.

this was in response (as in reaction, not followup) to:
https://sourceware.org/ml/libc-alpha/2018-05/msg00010.html

  if you want that change reverted, you will have to find someone else
  willing to do that; I won't.

-- 
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