On Mon, May 01, 2017 at 12:34:38PM +0200, Ævar Arnfjörð Bjarmason wrote:

quoted

I don't know if we would want to be extra paranoid about patch-ids.
There is no helping:

  git rev-list HEAD | git diff-tree --stdin -p | git patch-id --stable

because diff-tree doesn't know that it's trying for "--stable" output.
But the diffs we compute internally for patch-id could disable the
heuristics. I'm not sure if those matter, though. AFAIK those are used
only for internal comparisons within a single program. I.e., we never
compare them against input from the user, nor do we output them to the
user. So they'll change, but I don't think anybody would care.

I have a few-million row table with commit_id as one column & patch_id
as another. I.e. a commit -> patch_id mapping.

Thanks for this data point. It's always interesting to hear about
unforeseen uses of the tools.

Out of curiosity, how do you generate the patch-ids? Is it with
something like diff-tree piped to patch-id?

I do feel a bit sad about breaking this case (or at the very least
forcing you to set an option to retain cross-version compatibility). But
my gut says that we don't want to lock ourselves into never changing the
diff algorithm (and I'm sure we've done it inadvertently a few times
over the years; even the recent switch to turning on renames would have
had that impact).

-Peff