On Mon, May 1, 2023 at 9:59 AM Ævar Arnfjörð Bjarmason [off-list ref] wrote:

On Sat, Apr 22 2023, Elijah Newren via GitGitGadget wrote:

quoted

From: Elijah Newren <redacted>

Re earlier comments: If I rebase to make this the first commit
everything compiles, i.e. nothing here relied on the earlier split-offs
of cache.h into other headers.

You need to make a choice of whether to first split out cache.h, and
then do commits like these, or the other way around.

I'm not sure whether it's better to do it the other way around. If you
do that it's clear e.g. add-interactive.c's implicit dependency on
tree.h via commit.h has nothing to do with what would be the subsequent
split-up of cache.h.

Or maybe this is fine. I'm just trying to get some picture of what
depends on what in this series...

Yes, there is some freedom about the ordering of patches, and I had to
make a choice.

I found a number of the cleanups like this one for commit.h
interspersed with the other changes, but I intentionally grouped them
at the end to get a good high level overview, namely:

(A) Continue splitting declarations from cache.h to separate headers
(B) Do other header cleanups found during that work

I could have reversed the order, but since the series was motivated
and organized around (A), it made more sense to me to put those
patches first.

Besides, the last time I moved a few miscellaneous cleanup patches to
the front of the series, someone else responded thinking the purpose
and motivation was about those first few patches[1], so I wanted to
avoid a repeat of that problem.  ;-)

[1] https://lore.kernel.org/git/CABPp-BFZBWTG1VF6N8teVMYxoUdOeciKGwPq1g-G1K5--My5uQ@mail.gmail.com/ (local)