git/list[1] front-page[2] threads[3] people[4] search[5] about
 

[PATCH v2] rebase --autosquash: fix a potential segfault

From
Johannes Schindelin via GitGitGadget <gitgitgadget@gmail.com>
Date
May 5, 2020, 22:33 UTC
Message-ID
<pull.625.v2.git.1588717980225.gitgitgadget@gmail.com>
In-Reply-To
<pull.625.git.1588624804554.gitgitgadget@gmail.com>
From: Johannes Schindelin <johannes.schindelin@gmx.de>

When rearranging the todo list so that the fixups/squashes are reordered just after the commits they intend to fix up, we use two arrays to maintain that list: `next` and `tail`.

The idea is that `next[i]`, if set to a non-negative value, contains the index of the item that should be rearranged just after the `i`th item.

To avoid having to walk the entire `next` chain when appending another fixup/squash, we also store the end of the `next` chain in `tail[i]`.

The logic we currently use to update these array items is based on the assumption that given a fixup/squash item at index `i`, we just found the index `i2` indicating the first item in that fixup chain.

However, as reported by Paul Ganssle, that need not be true: the special form `fixup! <commit-hash>` is allowed to point to _another_ fixup commit in the middle of the fixup chain.

Example:
	* 0192a To fixup
	* 02f12 fixup! To fixup
	* 03763 fixup! To fixup
	* 04ecb fixup! 02f12

Note how the fourth commit targets the second commit, which is already a fixup that targets the first commit.

The good news is that it is easy to fix this: we use the correct condition (we now possibly set `tail[i2]` even for fixups in the middle) and we _also_ have to ensure that we _insert_ the item rather than _append_ it, i.e. we need to set `next[i2]` accordingly (it might still be set to `-1` if it was actually appended).

Reported-by: Paul Ganssle <paul@ganssle.io>
Helped-by: Jeff King <peff@peff.net>
Signed-off-by: Johannes Schindelin <johannes.schindelin@gmx.de>
---
    rebase --autosquash: fix a potential segfault
    
    This patch is in response to 
    https://lore.kernel.org/git/017dbc40-8d21-00fb-7b0e-6708d2dcb366@ganssle.io/
    .
    
    Changes since v1:
    
     * Fixed the order of two or more fixups-of-fixups (oddly enough, this
       simplified the patch...)
     * Extended the test to have two fixups-of-fixups, ensuring their order
       is preserved.
Published-As: https://github.com/gitgitgadget/git/releases/tag/pr-625%2Fdscho%2Fautosquash-segfault-v2
Fetch-It-Via: git fetch https://github.com/gitgitgadget/git pr-625/dscho/autosquash-segfault-v2
Pull-Request: https://github.com/gitgitgadget/git/pull/625
Range-diff vs v1:
 1:  bb820acc342 ! 1:  de029422324 rebase --autosquash: fix a potential segfault
     @@ Commit message
          index of the item that should be rearranged just after the `i`th item.
      
          To avoid having to walk the entire `next` chain when appending another
     -    fixup/squash, we also store the end of the `next` chain in `last[i]`.
     +    fixup/squash, we also store the end of the `next` chain in `tail[i]`.
      
          The logic we currently use to update these array items is based on the
          assumption that given a fixup/squash item at index `i`, we just found
     @@ Commit message
          Note how the fourth commit targets the second commit, which is already a
          fixup that targets the first commit.
      
     -    The good news is that it is easy to fix this: we can detect the
     -    situation by looking at `last[i2]` (which will be `-1` if `i2` is
     -    actually in the middle of a fixup chain), and in that case we simply
     -    need to squeeze the current item into the middle of the `next` chain,
     -    without touching `last` (i.e. leaving the end index of the fixup chain
     -    alone).
     +    The good news is that it is easy to fix this: we use the correct
     +    condition (we now possibly set `tail[i2]` even for fixups in the middle)
     +    and we _also_ have to ensure that we _insert_ the item rather than
     +    _append_ it, i.e. we need to set `next[i2]` accordingly (it might still
     +    be set to `-1` if it was actually appended).
      
          Reported-by: Paul Ganssle <paul@ganssle.io>
          Helped-by: Jeff King <peff@peff.net>
     @@ Commit message
      
       ## sequencer.c ##
      @@ sequencer.c: int todo_list_rearrange_squash(struct todo_list *todo_list)
     + 			todo_list->items[i].command =
     + 				starts_with(subject, "fixup!") ?
       				TODO_FIXUP : TODO_SQUASH;
     - 			if (next[i2] < 0)
     +-			if (next[i2] < 0)
     ++			if (tail[i2] < 0) {
     ++				next[i] = next[i2];
       				next[i2] = i;
      -			else
     -+			else if (tail[i2] >= 0)
     ++			} else {
     ++				next[i] = next[tail[i2]];
       				next[tail[i2]] = i;
     -+			else {
     -+				/*
     -+				 * i2 refers to a fixup commit in the middle of
     -+				 * a fixup chain
     -+				 */
     -+				next[i] = next[i2];
     -+				next[i2] = i;
     -+				continue;
      +			}
       			tail[i2] = i;
       		} else if (!hashmap_get_from_hash(&subject2item,
     @@ t/t3415-rebase-autosquash.sh: test_expect_success 'abort last squash' '
      +	git commit --squash HEAD^ -m Y --allow-empty &&
      +	test_tick &&
      +	git commit -m "squash! $(git rev-parse HEAD^)" -m Z --allow-empty &&
     -+	git rebase -ki --autosquash HEAD~4 &&
     -+	test XZY = $(git show | tr -cd X-Z)
     ++	test_tick &&
     ++	git commit -m "squash! $(git rev-parse HEAD^^)" -m W --allow-empty &&
     ++	git rebase -ki --autosquash HEAD~5 &&
     ++	test XZWY = $(git show | tr -cd W-Z)
      +'
      +
       test_done
 sequencer.c                  |  7 +++++--
 t/t3415-rebase-autosquash.sh | 16 ++++++++++++++++
 2 files changed, 21 insertions(+), 2 deletions(-)
diff --git a/sequencer.c b/sequencer.c
index e528225e787..d579f6d6c06 100644
--- a/sequencer.c
+++ b/sequencer.c
@@ -5264,10 +5264,13 @@ int todo_list_rearrange_squash(struct todo_list *todo_list)
 			todo_list->items[i].command =
 				starts_with(subject, "fixup!") ?
 				TODO_FIXUP : TODO_SQUASH;
-			if (next[i2] < 0)
+			if (tail[i2] < 0) {
+				next[i] = next[i2];
 				next[i2] = i;
-			else
+			} else {
+				next[i] = next[tail[i2]];
 				next[tail[i2]] = i;
+			}
 			tail[i2] = i;
 		} else if (!hashmap_get_from_hash(&subject2item,
 						strhash(subject), subject)) {
diff --git a/t/t3415-rebase-autosquash.sh b/t/t3415-rebase-autosquash.sh
index 093de9005b7..7bab6000dc7 100755
--- a/t/t3415-rebase-autosquash.sh
+++ b/t/t3415-rebase-autosquash.sh
@@ -424,4 +424,20 @@ test_expect_success 'abort last squash' '
 	! grep first actual
 '
 
+test_expect_success 'fixup a fixup' '
+	echo 0to-fixup >file0 &&
+	test_tick &&
+	git commit -m "to-fixup" file0 &&
+	test_tick &&
+	git commit --squash HEAD -m X --allow-empty &&
+	test_tick &&
+	git commit --squash HEAD^ -m Y --allow-empty &&
+	test_tick &&
+	git commit -m "squash! $(git rev-parse HEAD^)" -m Z --allow-empty &&
+	test_tick &&
+	git commit -m "squash! $(git rev-parse HEAD^^)" -m W --allow-empty &&
+	git rebase -ki --autosquash HEAD~5 &&
+	test XZWY = $(git show | tr -cd W-Z)
+'
+
 test_done

base-commit: af6b65d45ef179ed52087e80cb089f6b2349f4ec
-- 
gitgitgadget
Previous: Johannes SchindelinNext: Johannes Schindelin via GitGitGadget
Message 9 of 23 in “rebase --autosquash: fix a potential segfault”
  1. rebase --autosquash: fix a potential segfaultJohannes Schindelin via GitGitGadget, May 4, 2020
  2. Junio C HamanoMay 4, 2020
  3. Jeff KingMay 4, 2020
  4. Taylor BlauMay 4, 2020
  5. Junio C HamanoMay 5, 2020
  6. Johannes SchindelinMay 6, 2020
  7. Jeff KingMay 7, 2020
  8. Johannes SchindelinMay 8, 2020
  9. rebase --autosquash: fix a potential segfaultJohannes Schindelin via GitGitGadget, May 5, 2020
  10. rebase --autosquash: fix a potential segfaultJohannes Schindelin via GitGitGadget, May 9, 2020
  11. Andrei RybakMay 6, 2020
  12. Johannes SchindelinMay 7, 2020
  13. Philip OakleyMay 8, 2020
  14. Andrei RybakMay 8, 2020
  15. Philip OakleyMay 8, 2020
  16. Philip OakleyMay 18, 2020
  17. Johannes SchindelinMay 18, 2020
  18. Philip OakleyMay 25, 2020
  19. 0/2 Clarify some of the fixup! documenationPhilip Oakley, May 25, 2020
  20. 2/2 doc: fixup/squash: remove ellipsis marks, use <line> for clarifyPhilip Oakley, May 25, 2020
  21. 1/2 doc: fixup/squash: clarify use of <oid-hash> in subject linePhilip Oakley, May 25, 2020
  22. Junio C HamanoMay 27, 2020
  23. Philip OakleyMay 29, 2020

Read the whole thread, see it on lore, or plain text.

$ cat FOOTERMessages come from the public archive at lore.kernel.org/git, fetched every hour. The front page is chosen and written each morning by an AI editor and can be wrong; the threads themselves are the record. About and API. For agents: an MCP server at https://gitlist.dev/mcp, and any thread, story or person page as Markdown by adding .md to its URL (or sending Accept: text/markdown). Details in /llms.txt.