From: Jeff King <peff@peff.net>
Date: Fri, 10 Jun 2016 08:31:02 GMT
Subject: Re: [BUG-ish] diff compaction heuristic false positive
Message-ID: <20160610083102.GA14192@sigill.intra.peff.net>
In-Reply-To: <20160610075043.GA13411@sigill.intra.peff.net>

On Fri, Jun 10, 2016 at 03:50:43AM -0400, Jeff King wrote:

> I found a false positive with the new compaction heuristic in v2.9:
> [...]

And by the way, this is less "hey neat, I found a case" and more "wow,
this is a lot worse than I thought".

I diffed the old and new output for the top 10,000 commits in this
particular ruby code base. There were 45 commits with changed diffs.
Spot-checking them manually, a little over 1/3 of them featured this bad
pattern. The others looked like strict improvements.

That's a lot worse than the outcomes we saw on other code bases earlier.
1/3 bad is still a net improvement, so I dunno. Is this worth worrying
about? Should we bring back the documentation for the knob to disable
it? Should we consider making it tunable via gitattributes?

I don't think that last one really helps; the good cases _and_ the bad
ones are both in ruby code (though certainly the C code we looked at
earlier was all good).

It may also be possible to make it Just Work by using extra information
like indentation. I haven't thought hard enough about that to say.

-Peff