Triggered by one of brian's SHA-256 patch series, I looked at the reason why
the SHA-1 collision test case passed when it shouldn't. Turns out that the
regression test was not quite thorough enough, and the interactive rebase
did regress recently.

While in the area, I realized that the same bug exists in the code backing
the rebase.missingCommitsCheck feature: the backed-up todo list uses
shortened commit IDs that may very well become ambiguous during the rebase.
For good measure, this patch series fixes that, too.

Finally, I saw that git rebase --edit-todo reported the line in an awkward,
maybe even incorrect, way when there was an ambiguous commit ID, and I also
fixed that.

To make sure that the code can be easily adapted to SHA-256 after these
patches, I actually already made those adjustments on top and offered them
up at https://github.com/bk2204/git/pull/1.

Changes since v2:

 * Clarified in the first commit message that the change from a positive
   return value to a negative return value (to indicate failure, in both
   cases) both was intentional, and does not require any change in the
   corresponding function's only caller.

Changes since v1:

 * Turned the error condition when parsing the todo list with just-expanded
   commit IDs failed into a BUG().

Johannes Schindelin (3):
  parse_insn_line(): improve error message when parsing failed
  rebase -i: re-fix short SHA-1 collision
  rebase -i: also avoid SHA-1 collisions with missingCommitsCheck

 rebase-interactive.c          |  8 +++++---
 sequencer.c                   | 18 ++++++++++++++----
 t/t3404-rebase-interactive.sh | 17 +++++++++++++++--
 3 files changed, 34 insertions(+), 9 deletions(-)

base-commit: d0654dc308b0ba76dd8ed7bbb33c8d8f7aacd783

Submitted-As: https://lore.kernel.org/git/pull.529.v3.git.1579782499.gitgitgadget@gmail.com
In-Reply-To: https://lore.kernel.org/git/pull.529.git.1579209506.gitgitgadget@gmail.com
In-Reply-To: https://lore.kernel.org/git/pull.529.v2.git.1579304283.gitgitgadget@gmail.com