[RFC/PATCH,00/12] make "mktag" use fsck_tag()

Message

Ævar Arnfjörð Bjarmason Nov. 26, 2020, 1:28 a.m. UTC

This is a replacement for my earlier one-off SHA-1->SHA-256 cleanup
patch, on the basis of Jeff's comment that we should just be using
fsck_tag().

I was pretty much ready to send this as a non-RFC, but as seen in
12/12 fully duplicating the mktag logic in the context of using
fsck_tag() is a bit tricker than it looks. That breakage was only
revealed by a test I added while writing this.

The result is still nicer than the current code, but even then this
series breaks the tests mid-way through (in "improve verify_object()
test coverage"), but fixes them in 12/12.

I ran out of time to make this more sensible, but wanted to send an
RFC with the current state, and in case anyone else has any better
ideas for how to deal with 12/12. This could be applied as a non-RFC
by squashing 12/12 into the "use fsck instead of custom verify_tag"
change.

Ævar Arnfjörð Bjarmason (12):
  mktag: use default strbuf_read() hint
  mktag: reword write_object_file() error
  mktag: remove redundant braces in one-line body "if"
  mktag tests: don't needlessly use a subshell
  mktag tests: remove needless SHA-1 hardcoding
  mktag tests: improve verify_object() test coverage
  fsck: add new "extra" checks for "mktag"
  mktag: use fsck instead of custom verify_tag()
  mktag: remove now-unused verify_tag() code
  mktag doc: update to explain why to use this
  mktag docs: say <hash> not <sha1>
  mktag: bring back some of the verify_object() logic

 Documentation/git-hash-object.txt |   4 +
 Documentation/git-mktag.txt       |  31 ++++-
 builtin/mktag.c                   | 207 ++++++++++--------------------
 fsck.c                            |  32 ++++-
 fsck.h                            |   2 +
 t/t1006-cat-file.sh               |   2 +-
 t/t3800-mktag.sh                  | 103 ++++++++++-----
 7 files changed, 202 insertions(+), 179 deletions(-)