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

Re: [PATCH 22/22] lockfile: allow new file contents to be written while retaining lock

From
Jeff King <peff@peff.net>
Date
Apr 1, 2014, 20:39 UTC
Message-ID
<20140401203900.GI21715@sigill.intra.peff.net>
In-Reply-To
<1396367910-7299-23-git-send-email-mhagger@alum.mit.edu>
On Tue, Apr 01, 2014 at 05:58:30PM +0200, Michael Haggerty wrote:
Show 10 quoted lines
> Add a new option flag, LOCK_SEPARATE_STAGING_FILE, that can be passed
> to hold_lock_file_for_update() or hold_lock_file_for_append() to use a
> staging file that is independent of the lock file.
> 
> Add a new function activate_staging_file() that activates the contents
> that have been written to the staging file without releasing the lock.
> 
> This functionality can be used to ensure that changes to two files are
> seen by other processes in one order even if correctness requires the
> locks to be released in another order.

Can you give an example of when this is useful? I'm guessing the application is for writing out packed-refs before pruning loose refs in git-pack-refs?

It seems like this makes the API much more confusing. If I understand correctly, this is basically allowing us to take a lock, write to _another_ tmpfile that is not the lock, then rename the tmpfile into place without releasing the lock (and then we can drop the lock at our convenience).

I wonder if it would be simpler to build an API for that around the lock_file API, rather than as part of it. Or am I misunderstanding what's going on?

-Peff
Previous: Michael HaggertyNext: Eric Sunshine
Message 58 of 62 in “Lockfile refactoring and pre-activation”
  1. 00/22 Lockfile refactoring and pre-activationMichael Haggerty, Apr 1, 2014
  2. 01/22 t3204: test deleting references when lock files already existMichael Haggerty, Apr 1, 2014
  3. Jeff KingApr 1, 2014
  4. Michael HaggertyApr 2, 2014
  5. 02/22 try_merge_strategy(): remove redundant lock_file allocationMichael Haggerty, Apr 1, 2014
  6. Jeff KingApr 1, 2014
  7. Michael HaggertyApr 2, 2014
  8. Junio C HamanoApr 2, 2014
  9. Michael HaggertyApr 3, 2014
  10. 03/22 rollback_lock_file(): do not clear filename redundantlyMichael Haggerty, Apr 1, 2014
  11. 04/22 rollback_lock_file(): set fd to -1Michael Haggerty, Apr 1, 2014
  12. Jeff KingApr 1, 2014
  13. Junio C HamanoApr 2, 2014
  14. Michael HaggertyApr 6, 2014
  15. Junio C HamanoApr 7, 2014
  16. 05/22 lockfile: unlock file if lockfile permissions cannot be adjustedMichael Haggerty, Apr 1, 2014
  17. Jeff KingApr 1, 2014
  18. Jeff KingApr 1, 2014
  19. Torsten BögershausenApr 2, 2014
  20. Michael HaggertyApr 6, 2014
  21. 06/22 hold_lock_file_for_append(): release lock on errorsMichael Haggerty, Apr 1, 2014
  22. 07/22 lock_file(): always add lock_file object to lock_file_listMichael Haggerty, Apr 1, 2014
  23. Jeff KingApr 1, 2014
  24. Junio C HamanoApr 2, 2014
  25. Michael HaggertyApr 6, 2014
  26. Jeff KingApr 7, 2014
  27. 08/22 struct lock_file: replace on_list field with flags fieldMichael Haggerty, Apr 1, 2014
  28. 09/22 api-lockfile: expand the documentationMichael Haggerty, Apr 1, 2014
  29. Jeff KingApr 1, 2014
  30. Michael HaggertyApr 2, 2014
  31. 10/22 lockfile.c: document the various states of lock_file objectsMichael Haggerty, Apr 1, 2014
  32. 11/22 lockfile: define a constant LOCK_SUFFIX_LENMichael Haggerty, Apr 1, 2014
  33. Junio C HamanoApr 2, 2014
  34. 12/22 delete_ref_loose(): don't muck around in the lock_file's filenameMichael Haggerty, Apr 1, 2014
  35. Jeff KingApr 1, 2014
  36. Michael HaggertyApr 2, 2014
  37. Torsten BögershausenApr 2, 2014
  38. Jeff KingApr 2, 2014
  39. 13/22 config: change write_error() to take a (struct lock_file *) argumentMichael Haggerty, Apr 1, 2014
  40. Torsten BögershausenApr 2, 2014
  41. Michael HaggertyApr 6, 2014
  42. Junio C HamanoApr 2, 2014
  43. 14/22 lockfile: use strbufs when handling (most) pathsMichael Haggerty, Apr 1, 2014
  44. Jeff KingApr 1, 2014
  45. Junio C HamanoApr 2, 2014
  46. 15/22 resolve_symlink(): use a strbuf internallyMichael Haggerty, Apr 1, 2014
  47. 16/22 commit_lock_file(): don't work with a fixed-length bufferMichael Haggerty, Apr 1, 2014
  48. 17/22 lock_file(): exit early if lockfile cannot be openedMichael Haggerty, Apr 1, 2014
  49. 18/22 lockfile: also keep track of the filename of the file being lockedMichael Haggerty, Apr 1, 2014
  50. Junio C HamanoApr 2, 2014
  51. Michael HaggertyApr 6, 2014
  52. 19/22 struct lock_file: rename lock_filename field to staging_filenameMichael Haggerty, Apr 1, 2014
  53. 20/22 remove_lock_file(): call rollback_lock_file()Michael Haggerty, Apr 1, 2014
  54. 21/22 lockfile: extract a function reset_lock_file()Michael Haggerty, Apr 1, 2014
  55. Eric SunshineApr 2, 2014
  56. Michael HaggertyApr 2, 2014
  57. 22/22 lockfile: allow new file contents to be written while retaining lockMichael Haggerty, Apr 1, 2014
  58. Jeff KingApr 1, 2014
  59. Eric SunshineApr 2, 2014
  60. Junio C HamanoApr 2, 2014
  61. Jeff KingApr 1, 2014
  62. Michael HaggertyApr 3, 2014

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.