From: Derrick Stolee <stolee@gmail.com>
Date: Mon, 20 Aug 2018 18:14:20 GMT
Subject: Re: What's cooking in git.git (Aug 2018, #04; Fri, 17)
Message-ID: <800850d1-1880-2a8a-f5d5-7d0593787142@gmail.com>
In-Reply-To: <xmqqlg91aqzh.fsf@gitster-ct.c.googlers.com>

On 8/20/2018 1:26 PM, Junio C Hamano wrote:
> Jonathan Nieder <jrnieder@gmail.com> writes:
>
>>> Usually, I refrain from merging larger topics in 'next' down to
>>> 'master' when we get close to -rc0, but I am wondering if it is
>>> better to merge all of them to 'master', even the ones on the larger
>>> and possibly undercooked side, expecting that we collectively spend
>>> effort on hunting and fixing bugs in them during the pre-release
>>> freeze period.  If we were to go that route, I'd want everybody's
>>> buy-in and I'll promise to ignore any shiny new toys that appear on
>>> list that are not regression fixes to topics merged to 'master'
>>> since the end of the previous cycle to make sure people are not
>>> distracted.
>> Based on what I see in 'next' (midx, range-diff, etc), I quite like
>> this idea.
> The comment above was about the ones that was marked as "Will merge
> to 'master'" that are in 'next', not the ones marked as "Will cook
> in 'next'", like the midx ones.
>
> I am not worried about range-diff, as I do not think it came close
> enough to the core-ish code to break other things; the potential
> damage is fairly isolated and the worst would be that we'd realize
> that we shipped with a new command that was buggy after the release,
> which is not the end of the world.  'midx' and 'reachable' are quite
> different stories, as bugs in them would in the worst case lead to
> an immediate and unrecoverable repository corruption.
>
> So I am still on the fence, even though I am leaning toward merging
> them down to 'master'.

I'm happy to wait until 2.19 is cut for ds/multi-pack-index to merge 
down. I sent a new series today [1] that improves the series and 
corrects some bugs I found while making the full Git test suite work 
when writing a multi-pack-index on every repack. Nothing was 
"unrecoverable repository corruption," but I understand the concern 
there. Making the safest decision is probably the best decision.

I'm happy to contribute to a "pre-2.19 bug bash" after you merge things 
into master.

Thanks,

-Stolee

[1] 
https://public-inbox.org/git/20180820165124.152146-1-dstolee@microsoft.com/T/#u

     [PATCH 0/9] multi-pack-index cleanups