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

Re: [PATCH v4 1/5] setup: add an escape hatch for "no more default hash algorithm" change

From
Patrick Steinhardt <ps@pks.im>
Date
May 15, 2024, 12:23 UTC
Message-ID
<ZkSpXnsyxEf6p_4R@tanuki>
In-Reply-To
<xmqqikzg48hm.fsf@gitster.g>
On Tue, May 14, 2024 at 10:19:01AM -0700, Junio C Hamano wrote:
Show 27 quoted lines
> Junio C Hamano <gitster@pobox.com> writes:
> 
> > Partially revert c8aed5e8 (repository: stop setting SHA1 as the
> > default object hash, 2024-05-07), to keep end-user systems still
> > broken when we have gap in our test coverage but yet give them an
> > escape hatch to set the GIT_DEFAULT_HASH environment variable to
> > "sha1" in order to revert to the previous behaviour.  This variable
> > has been in use for using SHA-256 hash by default, and it should be
> > a better fit than inventing a new and test-only knob.
> 
> Having done all of this, I actually am very tempted to add the
> "always default to SHA-1" back as a fallback position to the
> set_default_hash_algo() function.  We know we are going to get the
> right hash algorithm when working in the repository, so the only
> case the default matters in practice is when working outside the
> repository.
> 
> We already have such a custom code for "git diff --no-index", and we
> are adding a few more back in here, but they can disappear if we had
> code to set the fallback default when GIT_DEFAULT_HASH does not
> exist here.  The "always use SHA-1 regardless of the hash used by
> the repository" code like "patch-id" should not depend on such a
> fallback default but should have its own code to explicitly set it.
> 
> As the user can tweak what algorithm they want if the wanted to, it
> does not sound too bad to have a fallback default when the user did
> not choose.

I think that this is going into the wrong direction. The patches that we have added surface real issues. If we now unconditionally add back the fallback to SHA-1, then we only punt those issues further down the road to the point in time where we drop `the_hash_algo` and/or `the_repository`. All the issues that were surfaced until now are technical debt, and the proposed fixes have been documented with a "TODO" item that they need more work.

In some cases it's as simple as adding in an option to the respective command that lets the user pick the hash algorithm, and I do not think that GIT_DEFAULT_HASH is a proper replacement for that. In other cases like for git-patch-id(1) the issue runs deeper and we need to refactor the whole subsystem to stop relying on `the_hash_algo` altogether.

Patrick
Previous: Junio C HamanoNext: Junio C Hamano
Message 35 of 62 in “Fix use of uninitialized hash algos”
  1. 0/2 Fix use of uninitialized hash algosPatrick Steinhardt, May 13, 2024
  2. 1/2 builtin/patch-id: fix uninitialized hash functionPatrick Steinhardt, May 13, 2024
  3. 2/2 builtin/hash-object: fix uninitialized hash functionPatrick Steinhardt, May 13, 2024
  4. Junio C HamanoMay 14, 2024
  5. Junio C HamanoMay 13, 2024
  6. Junio C HamanoMay 13, 2024
  7. 0/4 Fix use of uninitialized hash algorithmsJunio C Hamano, May 13, 2024
  8. 1/4 setup: add an escape hatch for "no more default hash algorithm" changeJunio C Hamano, May 13, 2024
  9. Kyle LippincottMay 13, 2024
  10. 2/4 t1517: test commands that are designed to be run outside repositoryJunio C Hamano, May 13, 2024
  11. Kyle LippincottMay 13, 2024
  12. Junio C HamanoMay 13, 2024
  13. Junio C HamanoMay 13, 2024
  14. Kyle LippincottMay 13, 2024
  15. 3/4 builtin/patch-id: fix uninitialized hash functionJunio C Hamano, May 13, 2024
  16. 4/4 builtin/hash-object: fix uninitialized hash functionJunio C Hamano, May 13, 2024
  17. 5/4 apply: fix uninitialized hash functionJunio C Hamano, May 13, 2024
  18. 0/5 Fix use of uninitialized hash algorithmsJunio C Hamano, May 13, 2024
  19. 1/5 setup: add an escape hatch for "no more default hash algorithm" changeJunio C Hamano, May 13, 2024
  20. 2/5 t1517: test commands that are designed to be run outside repositoryJunio C Hamano, May 13, 2024
  21. 3/5 builtin/patch-id: fix uninitialized hash functionJunio C Hamano, May 13, 2024
  22. Junio C HamanoMay 13, 2024
  23. Patrick SteinhardtMay 14, 2024
  24. Junio C HamanoMay 14, 2024
  25. 4/5 builtin/hash-object: fix uninitialized hash functionJunio C Hamano, May 13, 2024
  26. Junio C HamanoMay 13, 2024
  27. Patrick SteinhardtMay 14, 2024
  28. Junio C HamanoMay 14, 2024
  29. 5/5 apply: fix uninitialized hash functionJunio C Hamano, May 13, 2024
  30. 0/5 Fix use of uninitialized hash algorithmsJunio C Hamano, May 14, 2024
  31. 1/5 setup: add an escape hatch for "no more default hash algorithm" changeJunio C Hamano, May 14, 2024
  32. Patrick SteinhardtMay 14, 2024
  33. Junio C HamanoMay 14, 2024
  34. Junio C HamanoMay 14, 2024
  35. Patrick SteinhardtMay 15, 2024
  36. Junio C HamanoMay 16, 2024
  37. 2/5 t1517: test commands that are designed to be run outside repositoryJunio C Hamano, May 14, 2024
  38. Patrick SteinhardtMay 14, 2024
  39. Junio C HamanoMay 14, 2024
  40. Patrick SteinhardtMay 15, 2024
  41. Junio C HamanoMay 15, 2024
  42. Patrick SteinhardtMay 15, 2024
  43. Junio C HamanoMay 15, 2024
  44. 3/5 builtin/patch-id: fix uninitialized hash functionJunio C Hamano, May 14, 2024
  45. 4/5 builtin/hash-object: fix uninitialized hash functionJunio C Hamano, May 14, 2024
  46. Junio C HamanoMay 17, 2024
  47. Junio C HamanoMay 20, 2024
  48. Junio C HamanoMay 20, 2024
  49. 5/5 apply: fix uninitialized hash functionJunio C Hamano, May 14, 2024
  50. 0/5 Fix use of uninitialized hash algorithmsJunio C Hamano, May 20, 2024
  51. 1/5 setup: add an escape hatch for "no more default hash algorithm" changeJunio C Hamano, May 20, 2024
  52. Patrick SteinhardtMay 21, 2024
  53. Junio C HamanoMay 21, 2024
  54. 2/5 t1517: test commands that are designed to be run outside repositoryJunio C Hamano, May 20, 2024
  55. 3/5 builtin/patch-id: fix uninitialized hash functionJunio C Hamano, May 20, 2024
  56. 4/5 builtin/hash-object: fix uninitialized hash functionJunio C Hamano, May 20, 2024
  57. 5/5 apply: fix uninitialized hash functionJunio C Hamano, May 20, 2024
  58. Patrick SteinhardtMay 21, 2024
  59. Junio C HamanoMay 21, 2024
  60. Patrick SteinhardtMay 21, 2024
  61. Junio C HamanoMay 21, 2024
  62. Patrick SteinhardtMay 22, 2024

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.