Re: [PATCH] clean: demonstrate a bug with pathspecs

On Wed, Jan 15, 2020 at 5:23 PM Derrick Stolee [off-list ref] wrote:

On 1/15/2020 7:38 PM, Elijah Newren wrote:

quoted

Is there an inverted phrase corresponding to "the gift that keeps on
giving", something like "the punishment that keeps on punishing"?  If
so, it would be a very appropriate description of dir.c.

At least we will continue adding tests until we converge towards
correctness, and the behavior issues are even more contrived and
special case (like this one).

This doesn't seem any more contrived or special case than most my
previous fixes for dir.c...

quoted

Yeah, I still have context.  I even think I've got an idea about what
the fix might be, though with dir.c my ideas about fixes usually just
serve as starting points for debugging before I find the real fix.
I'll try to dig in.

Thanks! I'll try to review it carefully when it arrives. Good luck.

Man, I'm such a bozo.  It turns out, for once, that my idea for the
fix was correct but after digging a bit I realized that it was
essentially a bug I fixed not that long ago once already -- and that I
myself re-introduced it (for a slightly different case) in some
commits where I used some strongly worded disgust that "this bad code
structure is going to cause someone to mess up in <this way>" and then
I made that exact kind of mistake I was complaining about in the
commit message...as part of that EXACT commit, to boot.

At least it'll make for a fun new commit message explaining it all...


Anyway, I'm going to pull your commit into my series so I can put my
fix on top, and lump it in with Peff's two patches over at
https://lore.kernel.org/git/20200115202146.GA4091171@coredump.intra.peff.net/ (local)
since all these patches are basically "more fill_directory() fixes".
Let me know if you have any concerns with that.

Elijah