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

Re: [PATCH 00/28] Create a reference backend for packed refs

From
Michael Haggerty <mhagger@alum.mit.edu>
Date
Jun 22, 2017, 07:36 UTC
Message-ID
<05ea33a0-adde-54df-07e8-bdfba9c7d503@alum.mit.edu>
In-Reply-To
<20170619195330.fhjlfiqmcwhgttaa@sigill.intra.peff.net>
On 06/19/2017 09:53 PM, Jeff King wrote:
Show 17 quoted lines
> On Mon, Jun 19, 2017 at 03:43:15PM -0400, Jeff King wrote:
> 
>>>> Is the iterator over packed-refs correctly skipping over what are
>>>> covered by loose refs?  The entries in the packed-refs file that are
>>>> superseded by loose refs should be allowed to point at an already
>>>> expired object.
>>>
>>> Here it is in a test form for easier diagnosis.
>>
>> Thanks, I was just starting to do that myself. The problem is in
>> ca6b06eb7 (packed_ref_store: support iteration, 2017-05-15) and is
>> pretty obvious: the packed_ref iterator checks whether the entry
>> resolves.
>>
>> I think that _neither_ of the loose and packed iterators should be
>> checking this. It's only the merged result (where loose trumps packed)
>> that should bother checking.
Thanks for the bug report and the analysis, which is exactly right.

But I'd like to fix the problem a *little* differently than Peff suggested. To keep `packed_ref_store` from deviating more than necessary from the `ref_store` interface, I propose that we leave the code for rejecting broken refs where it is, and instead invoke `packed_ref_iterator_begin()` with the `DO_FOR_EACH_INCLUDE_BROKEN` flag.

I have prepared a re-roll of the patch series, but I can't submit it until I have Junio's signoff on the test that he suggested [1]. Junio?

Thanks, Michael

[1] http://public-inbox.org/git/xmqqvanrsru4.fsf@gitster.mtv.corp.google.com/

Previous: Jeff KingNext: Junio C Hamano
Message 51 of 52 in “Create a reference backend for packed refs”
  1. 00/28 Create a reference backend for packed refsMichael Haggerty, Jun 15, 2017
  2. 01/28 add_packed_ref(): teach function to overwrite existing refsMichael Haggerty, Jun 15, 2017
  3. 03/28 packed_ref_store: move `packed_refs_path` hereMichael Haggerty, Jun 15, 2017
  4. Stefan BellerJun 16, 2017
  5. 04/28 packed_ref_store: move `packed_refs_lock` member hereMichael Haggerty, Jun 15, 2017
  6. Stefan BellerJun 16, 2017
  7. Michael HaggertyJun 16, 2017
  8. Stefan BellerJun 16, 2017
  9. 07/28 get_packed_ref_cache(): take a `packed_ref_store *` parameterMichael Haggerty, Jun 15, 2017
  10. 08/28 get_packed_refs(): take a `packed_ref_store *` parameterMichael Haggerty, Jun 15, 2017
  11. 09/28 add_packed_ref(): take a `packed_ref_store *` parameterMichael Haggerty, Jun 15, 2017
  12. 10/28 lock_packed_refs(): take a `packed_ref_store *` parameterMichael Haggerty, Jun 15, 2017
  13. 11/28 commit_packed_refs(): take a `packed_ref_store *` parameterMichael Haggerty, Jun 15, 2017
  14. 13/28 get_packed_ref(): take a `packed_ref_store *` parameterMichael Haggerty, Jun 15, 2017
  15. 14/28 repack_without_refs(): take a `packed_ref_store *` parameterMichael Haggerty, Jun 15, 2017
  16. 15/28 packed_peel_ref(): new function, extracted from `files_peel_ref()`Michael Haggerty, Jun 15, 2017
  17. Stefan BellerJun 16, 2017
  18. Michael HaggertyJun 16, 2017
  19. Stefan BellerJun 16, 2017
  20. 16/28 packed_ref_store: support iterationMichael Haggerty, Jun 15, 2017
  21. 17/28 packed_read_raw_ref(): new function, replacing `resolve_packed_ref()`Michael Haggerty, Jun 15, 2017
  22. 19/28 packed_ref_store: make class into a subclass of `ref_store`Michael Haggerty, Jun 15, 2017
  23. 20/28 commit_packed_refs(): report errors rather than dyingMichael Haggerty, Jun 15, 2017
  24. 21/28 commit_packed_refs(): use a staging file separate from the lockfileMichael Haggerty, Jun 15, 2017
  25. Junio C HamanoJul 20, 2017
  26. Jonathan NiederJul 20, 2017
  27. Junio C HamanoJul 21, 2017
  28. Junio C HamanoJul 21, 2017
  29. Jeff KingJul 24, 2017
  30. Jonathan NiederJul 24, 2017
  31. Jeff KingJul 24, 2017
  32. Junio C HamanoJul 24, 2017
  33. Michael HaggertyJul 26, 2017
  34. 22/28 packed_refs_lock(): function renamed from lock_packed_refs()Michael Haggerty, Jun 15, 2017
  35. 18/28 packed-backend: new module for handling packed referencesMichael Haggerty, Jun 15, 2017
  36. Stefan BellerJun 16, 2017
  37. 24/28 packed_refs_unlock(), packed_refs_is_locked(): new functionsMichael Haggerty, Jun 15, 2017
  38. 26/28 commit_packed_refs(): remove call to `packed_refs_unlock()`Michael Haggerty, Jun 15, 2017
  39. 28/28 read_packed_refs(): die if `packed-refs` contains bogus dataMichael Haggerty, Jun 15, 2017
  40. 23/28 packed_refs_lock(): report errors via a `struct strbuf *err`Michael Haggerty, Jun 15, 2017
  41. 27/28 repack_without_refs(): don't lock or unlock the packed refsMichael Haggerty, Jun 15, 2017
  42. 25/28 clear_packed_ref_cache(): don't protest if the lock is heldMichael Haggerty, Jun 15, 2017
  43. 12/28 rollback_packed_refs(): take a `packed_ref_store *` parameterMichael Haggerty, Jun 15, 2017
  44. 02/28 packed_ref_store: new structMichael Haggerty, Jun 15, 2017
  45. 06/28 validate_packed_ref_cache(): take a `packed_ref_store *` parameterMichael Haggerty, Jun 15, 2017
  46. 05/28 clear_packed_ref_cache(): take a `packed_ref_store *` parameterMichael Haggerty, Jun 15, 2017
  47. Junio C HamanoJun 19, 2017
  48. Junio C HamanoJun 19, 2017
  49. Jeff KingJun 19, 2017
  50. Jeff KingJun 19, 2017
  51. Michael HaggertyJun 22, 2017
  52. Junio C HamanoJun 22, 2017

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.