[PATCH v7 0/5] [GSoC] remove_subtree(): reimplement using iterators

[Daniel Ferreira <bnmvco@gmail.com>]

This is the seventh version of a patch series that implements the GSoC
microproject of converting a recursive call to readdir() to use dir_iterator.

v1: https://public-inbox.org/git/CAGZ79kZwT-9mHTiOJ5CEjk2wDFkn6+NcogjX0=vjhsAh16ANYg@mail.gmail.com/T/#t
v2: https://public-inbox.org/git/CACsJy8Dxh-QPBBLfaFWPAWUsbA9GVXA7x+mXLjEvYKhk1zOpig@mail.gmail.com/T/#t
v3: https://public-inbox.org/git/CAGZ79kYtpmURSQWPumobA=e3JBFjKhWCdv_LPhKCd71ZRwMovA@mail.gmail.com/T/#t
v4: https://public-inbox.org/git/1490747533-89143-1-git-send-email-bnmvco@gmail.com/T/#e437a63e0c22c00c69b5d92977c9b438ed2b9fd3a
v5: https://public-inbox.org/git/1490844730-47634-1-git-send-email-bnmvco@gmail.com/T/#m2323f15e45de699f2e09364f40a62e17047cf453
v6: https://public-inbox.org/git/1491107726-21504-1-git-send-email-bnmvco@gmail.com/T/#t

I screwed up in v6 because I had introduced a bug that in case git tried to open
a directory that did not exist using dir_iterator, the program would segfault. This
was amended and all commits are passing the tests. Sorry for not having tested
my changes properly.

CI build: https://travis-ci.org/theiostream/git

Since the changes in v6 were not reviewed, I'll just copy what was sent
back there.

> Back in v5, Michael had a number of suggestions, all of which were applied
> to this version (including a slightly modified version of his "biggish rewrite"
> project to make dir_iterator's state machine simpler). The only suggestion that
> did not make it into this series was that of not traversing into subdirectories,
> since I believe it would be better off in another series that actually required
> that feature (that is, I do not want a series to implement a feature it will
> not need). The same goes for Junio's thought on a flag to list *only* directories
> and no files on the v4 discussion.

> Junio and Peff's comments about how to write to files in the tests were also
> considered, and the tests were adjusted.

> I chose to squash both the state machine refactor and the addition of the
> new flags in a single commit. I do not know whether you will feel this is
> the right choice but it seemed natural, since most of the state machine's
> new logic would not even make sense without encompassing the new features.
> I am, of course, open for feedback on this decision.

Daniel Ferreira (5):
  dir_iterator: add tests for dir_iterator API
  remove_subtree(): test removing nested directories
  dir_iterator: add helpers to dir_iterator_advance
  dir_iterator: refactor state machine model
  remove_subtree(): reimplement using iterators

 Makefile                        |   1 +
 dir-iterator.c                  | 196 ++++++++++++++++++++++++++--------------
 dir-iterator.h                  |  28 ++++--
 entry.c                         |  38 +++-----
 refs/files-backend.c            |   2 +-
 t/helper/.gitignore             |   1 +
 t/helper/test-dir-iterator.c    |  32 +++++++
 t/t0065-dir-iterator.sh         | 109 ++++++++++++++++++++++
 t/t2000-checkout-cache-clash.sh |  11 +++
 9 files changed, 316 insertions(+), 102 deletions(-)
 create mode 100644 t/helper/test-dir-iterator.c
 create mode 100755 t/t0065-dir-iterator.sh

--
2.7.4 (Apple Git-66)