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

Re: [PATCH 15/29] ref_transaction_create(): disallow recursive pruning

From
Junio C Hamano <gitster@pobox.com>
Date
Apr 29, 2016, 08:41 UTC
Message-ID
<xmqqzisc24en.fsf@gitster.mtv.corp.google.com>
In-Reply-To
<5723059C.5080406@alum.mit.edu>
Two things I forgot to say.
Michael Haggerty <mhagger@alum.mit.edu> writes:
> I was thinking of this patch as documenting and enforcing a limitation
> in the current implementation of pruning.

This actually is an excellent point, and is the reason why I repeatedly made my suggestion conditional on "If my understanding of the purpose of this patch correctly" ;-) And I personally am not convinced that this limitation is fundamental and will not be lifted. It may be better to treat this as a limitation of the current implementation of pruning, and in that case I do agree that the original patch that started this thread is the best way to express it.

I said that it makes harder to write incorrect client code of this API to make REF_ISPRUNING contain both 0x04 and REF_NODEREF, and I still think that is true, but at the same time, it makes it easier to write incorrect code in the API implementation. We no longer can check if we are in the pruning codepath with

    if (flag & REF_ISPRUNING)
but have to write
    if ((flag & REF_ISPRUNING) == REF_ISPRUNING)

instead. This "& with the constant and compare the result with that same constant" pattern could also be used for a single-bit constant, so if there were short-and-sweet syntax to express that pattern in the language, consistently using that for all bit checks would make it less likely for us to write incorrect code, but but there is no short-and-sweet syntax to do so in C, and spelling it out like the above is too cumbersome to be practical. The suggested squash did

    if (flag & REF_ISPRUNING_)

to check only for 0x04 bit, but that is also error prone; it is too easy to forget the difference between the two and drop the trailing underscore by mistake.

Even though it is generally a good trade-off to make it harder to make mistakes in the "user of the API" code even if it makes it easier to make mistakes in the "implementation of the API" code, because the "user of the API" in this case is actually still inside the ref API implementation, I do not think either way makes too much of a difference. So perhaps your original might be the best version among those that have been discussed in this thread.

Previous: Junio C HamanoNext: Michael Haggerty
Message 35 of 70 in “Yet more preparation for reference backends”
  1. 00/29 Yet more preparation for reference backendsMichael Haggerty, Apr 27, 2016
  2. 01/29 safe_create_leading_directories(): improve docstringMichael Haggerty, Apr 27, 2016
  3. 02/29 remove_dir_recursively(): add docstringMichael Haggerty, Apr 27, 2016
  4. 03/29 refname_is_safe(): use skip_prefix()Michael Haggerty, Apr 27, 2016
  5. 04/29 refname_is_safe(): don't allow the empty stringMichael Haggerty, Apr 27, 2016
  6. 05/29 refname_is_safe(): insist that the refname already be normalizedMichael Haggerty, Apr 27, 2016
  7. Junio C HamanoApr 27, 2016
  8. David TurnerApr 27, 2016
  9. Jeff KingApr 27, 2016
  10. David TurnerApr 27, 2016
  11. Jeff KingApr 27, 2016
  12. Jeff KingApr 27, 2016
  13. David TurnerApr 28, 2016
  14. 06/29 commit_ref_update(): write error message to *err, not stderrMichael Haggerty, Apr 27, 2016
  15. 07/29 rename_ref(): remove unneeded local variableMichael Haggerty, Apr 27, 2016
  16. 08/29 ref_transaction_commit(): remove local variable nMichael Haggerty, Apr 27, 2016
  17. Junio C HamanoApr 27, 2016
  18. Junio C HamanoApr 27, 2016
  19. 09/29 read_raw_ref(): rename flags argument to typeMichael Haggerty, Apr 27, 2016
  20. 10/29 read_raw_ref(): clear *type at start of functionMichael Haggerty, Apr 27, 2016
  21. 11/29 read_raw_ref(): rename symref argument to referentMichael Haggerty, Apr 27, 2016
  22. 12/29 read_raw_ref(): improve docstringMichael Haggerty, Apr 27, 2016
  23. Junio C HamanoApr 27, 2016
  24. 13/29 lock_ref_sha1_basic(): remove unneeded local variableMichael Haggerty, Apr 27, 2016
  25. 14/29 refs: make error messages more consistentMichael Haggerty, Apr 27, 2016
  26. 15/29 ref_transaction_create(): disallow recursive pruningMichael Haggerty, Apr 27, 2016
  27. Junio C HamanoApr 27, 2016
  28. David TurnerApr 27, 2016
  29. Junio C HamanoApr 27, 2016
  30. Junio C HamanoApr 27, 2016
  31. David TurnerApr 28, 2016
  32. Junio C HamanoApr 28, 2016
  33. Michael HaggertyApr 29, 2016
  34. Junio C HamanoApr 29, 2016
  35. Junio C HamanoApr 29, 2016
  36. Michael HaggertyApr 29, 2016
  37. 16/29 ref_transaction_commit(): correctly report close_ref() failureMichael Haggerty, Apr 27, 2016
  38. 17/29 delete_branches(): use resolve_refdup()Michael Haggerty, Apr 27, 2016
  39. 18/29 refs: allow log-only updatesMichael Haggerty, Apr 27, 2016
  40. 19/29 refs: don't dereference on renameMichael Haggerty, Apr 27, 2016
  41. Junio C HamanoApr 27, 2016
  42. Michael HaggertyApr 29, 2016
  43. Junio C HamanoApr 29, 2016
  44. Michael HaggertyApr 29, 2016
  45. Jeff KingApr 29, 2016
  46. Michael HaggertyApr 29, 2016
  47. Jeff KingApr 29, 2016
  48. Junio C HamanoApr 29, 2016
  49. David TurnerApr 29, 2016
  50. Michael HaggertyApr 30, 2016
  51. Junio C HamanoMay 2, 2016
  52. 20/29 verify_refname_available(): adjust constness in declarationMichael Haggerty, Apr 27, 2016
  53. 21/29 add_update(): initialize the whole ref_updateMichael Haggerty, Apr 27, 2016
  54. 22/29 lock_ref_for_update(): new functionMichael Haggerty, Apr 27, 2016
  55. 23/29 unlock_ref(): move definition higher in the fileMichael Haggerty, Apr 27, 2016
  56. 24/29 ref_transaction_update(): check refname_is_safe() at a minimumMichael Haggerty, Apr 27, 2016
  57. Junio C HamanoApr 27, 2016
  58. Michael HaggertyApr 29, 2016
  59. Junio C HamanoApr 29, 2016
  60. 25/29 refs: resolve symbolic refs firstMichael Haggerty, Apr 27, 2016
  61. David TurnerApr 28, 2016
  62. Michael HaggertyApr 29, 2016
  63. David TurnerApr 29, 2016
  64. Junio C HamanoMay 2, 2016
  65. 26/29 lock_ref_for_update(): don't re-read non-symbolic referencesMichael Haggerty, Apr 27, 2016
  66. 27/29 lock_ref_for_update(): don't resolve symrefsMichael Haggerty, Apr 27, 2016
  67. 28/29 commit_ref_update(): remove the flags parameterMichael Haggerty, Apr 27, 2016
  68. 29/29 lock_ref_sha1_basic(): only handle REF_NODEREF modeMichael Haggerty, Apr 27, 2016
  69. Junio C HamanoApr 29, 2016
  70. David TurnerApr 29, 2016

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.