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

Re: [PATCH v4 4/6] pack-objects: generate cruft packs at most one object over threshold

From
Junio C Hamano <gitster@pobox.com>
Date
Mar 11, 2025, 21:59 UTC
Message-ID
<xmqqikof2pqp.fsf@gitster.g>
In-Reply-To
<f2ca92245ada74825806b50f786aab312275fd85.1741648467.git.me@ttaylorr.com>
Taylor Blau <me@ttaylorr.com> writes:
Show 8 quoted lines
> When generating multiple cruft packs with 'git repack --max-cruft-size',
> we use 'git pack-objects --cruft --max-pack-size' (with many other
> elided options), filling in the '--max-pack-size' value with whatever
> was provided via the '--max-cruft-size' flag.
>
> This causes us to generate a pack that is smaller than the specified
> threshold. This poses a problem since we will never be able to generate
> a cruft pack that crosses the threshold.

So far I see absolutely *NO* problem described in the above. The user said "I want to chop them into 200MB pieces but do not exceed the threshold" and the system honored that wish.

> In effect, this means that we
> will try and repack its contents over and over again.

The end effect however may be problematic, but isn't it due to the way when to repack is determined? You see 199MB piece of cruft pack plus some other cruft data. You have generated no new cruft and no existing cruft expired out, but you do not know these facts until you try to repack. Because 200MB is the limit, you include the 199MB one as part of the ones to be recombined into the new cruft pack because 199MB is smaller than 200MB and you do not know that the reason why it is 199MB is because the earlier repack operation found all remaining cruft material to be larger than 1MB; if there were a 0.5MB cruft, it may have made it closer to 200MB.

So would it be feasible to remember how 199MB cruft pack is lying in the object store (i.e. earlier we packed as much as possible), and add a logic that says "if there is nothing to expire out of this one, do not attempt to repack---this is fine as-is"?

> Instead, change the meaning of '--max-pack-size' in pack-objects when
> combined with '--cruft'. When put together, '--max-pack-size' allows the
> pack to grow larger than the specified threshold, but only by one
> additional object.

I do not think that would work well. You have no control over the size of that one additional object---it may weigh more than 100MB, combining your 199MB cruft pack with something else to make it ~300MB cruft. In other words, "just a little bit larger" sounds like a wishful thinking handwaving.

Previous: Taylor BlauNext: Taylor Blau
Message 30 of 54 in “pack-objects: freshen objects with multi-cruft packs”
  1. 0/2 pack-objects: freshen objects with multi-cruft packsTaylor Blau, Feb 27, 2025
  2. 1/2 builtin/repack.c: simplify cruft pack aggregationTaylor Blau, Feb 27, 2025
  3. Elijah NewrenFeb 27, 2025
  4. Taylor BlauFeb 27, 2025
  5. Patrick SteinhardtFeb 28, 2025
  6. Elijah NewrenMar 4, 2025
  7. Junio C HamanoMar 5, 2025
  8. Taylor BlauMar 5, 2025
  9. 2/2 builtin/pack-objects.c: freshen objects from existing cruft packsTaylor Blau, Feb 27, 2025
  10. Elijah NewrenFeb 27, 2025
  11. Taylor BlauFeb 27, 2025
  12. Elijah NewrenFeb 27, 2025
  13. Taylor BlauFeb 27, 2025
  14. 0/2 pack-objects: freshen objects with multi-cruft packsTaylor Blau, Mar 4, 2025
  15. 1/2 builtin/repack.c: simplify cruft pack aggregationTaylor Blau, Mar 4, 2025
  16. 2/2 builtin/pack-objects.c: freshen objects from existing cruft packsTaylor Blau, Mar 4, 2025
  17. Elijah NewrenMar 4, 2025
  18. Taylor BlauMar 5, 2025
  19. Taylor BlauMar 5, 2025
  20. 0/1 pack-objects: freshen objects with multi-cruft packsTaylor Blau, Mar 5, 2025
  21. 1/1 builtin/pack-objects.c: freshen objects from existing cruft packsTaylor Blau, Mar 5, 2025
  22. Patrick SteinhardtMar 6, 2025
  23. Taylor BlauMar 13, 2025
  24. Patrick SteinhardtMar 6, 2025
  25. 0/6 pack-objects: freshen objects with multi-cruft packsTaylor Blau, Mar 11, 2025
  26. 1/6 t/t5329-pack-objects-cruft.sh: evict 'repack'-related testsTaylor Blau, Mar 11, 2025
  27. 2/6 t7704-repack-cruft.sh: consolidate `write_blob()`Taylor Blau, Mar 11, 2025
  28. 3/6 t/lib-cruft.sh: extract some cruft-related helpersTaylor Blau, Mar 11, 2025
  29. 4/6 pack-objects: generate cruft packs at most one object over thresholdTaylor Blau, Mar 11, 2025
  30. Junio C HamanoMar 11, 2025
  31. Taylor BlauMar 12, 2025
  32. Junio C HamanoMar 12, 2025
  33. Taylor BlauMar 12, 2025
  34. Elijah NewrenMar 12, 2025
  35. Taylor BlauMar 12, 2025
  36. Junio C HamanoMar 12, 2025
  37. Elijah NewrenMar 12, 2025
  38. Junio C HamanoMar 13, 2025
  39. Elijah NewrenMar 13, 2025
  40. Junio C HamanoMar 13, 2025
  41. 5/6 builtin/repack.c: simplify cruft pack aggregationTaylor Blau, Mar 11, 2025
  42. 6/6 builtin/pack-objects.c: freshen objects from existing cruft packsTaylor Blau, Mar 11, 2025
  43. Junio C HamanoMar 11, 2025
  44. Taylor BlauMar 12, 2025
  45. Junio C HamanoMar 12, 2025
  46. Taylor BlauMar 12, 2025
  47. Junio C HamanoMar 12, 2025
  48. Taylor BlauMar 12, 2025
  49. Junio C HamanoMar 13, 2025
  50. Taylor BlauMar 13, 2025
  51. Jeff KingMar 13, 2025
  52. Junio C HamanoMar 13, 2025
  53. builtin/pack-objects.c: freshen objects from existing cruft packsTaylor Blau, Mar 13, 2025
  54. Junio C HamanoMar 13, 2025

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.