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

Re: Bizarre missing changes (git bug?)

From
RZRoman Zippel <zippel@linux-m68k.org>
Date
Jul 26, 2008, 03:12 UTC
Message-ID
<200807260512.40088.zippel@linux-m68k.org>
In-Reply-To
<alpine.LFD.1.10.0807211331390.31863@woody.linux-foundation.org>
Hi,
On Monday 21. July 2008, Linus Torvalds wrote:
Show 12 quoted lines
> Read up on '--full-history'.
>
> By default, git simplifies the history for logs that have path
> simplification: only walking down the side of a merge that all the data
> came from (ie the unchanged side). So it only leaves merges around if
> there was changes from _all_ parents.
>
> So without --full-history, if any parent matches the state, it just
> removes the merge and picks that parent that contained all the state.
> Obviously, any changes to that file can be sufficiently explained by
> walking just that limited history, because they must have changed in
> _that_ history too!

Is that really a good default behaviour? Is there a way to change that default?

I'm currently looking into converting the m68k CVS repository and I'd like to properly regenerate it as two separate lines of development. The problem is if I look at the file history, I often only see one side of the changes when things got merged because of this default. What makes this worse is that graphical front ends may inherit this behaviour. I tested this with qgit and it only shows half of the history. giggle retrieves the history like --full-history, but it lacks empty merges, so it makes it harder to see what got merged when. For example a history that actually looks this:

linux -+-----import----+-----------import----+------... m68k \-commit-commit-\-merge-commit-commit-\-merge...

Without the merges it looks like this:

linux -+-----import----------------import--------------+... m68k \-commit-commit---------commit-commit \...

And without --full-history these "loose ends" aren't visible as in qgit.

When researching historical changes one wants to know when something was introduced and when it was merged, but this simplification makes it harder than it really has to be. IMO the default should be to show the complete history, so one doesn't miss something by accident that might be important or as the original example shows it might be confusing if one sees a change with "git log --stat id.." and the change disappears when one looks at "git log path".

bye, Roman
Previous: Tim HarperNext: Linus Torvalds
Message 5 of 58 in “Bizarre missing changes (git bug?)”
  1. Tim HarperJul 21, 2008
  2. Linus TorvaldsJul 21, 2008
  3. Tim HarperJul 21, 2008
  4. Tim HarperJul 21, 2008
  5. Roman ZippelJul 26, 2008
  6. Linus TorvaldsJul 26, 2008
  7. Roman ZippelJul 27, 2008
  8. Linus TorvaldsJul 27, 2008
  9. Roman ZippelJul 27, 2008
  10. Linus TorvaldsJul 27, 2008
  11. Roman ZippelJul 28, 2008
  12. Linus TorvaldsJul 28, 2008
  13. Linus TorvaldsJul 28, 2008
  14. Roman ZippelJul 29, 2008
  15. Martin LanghoffJul 29, 2008
  16. Roman ZippelJul 30, 2008
  17. Martin LanghoffJul 30, 2008
  18. Linus TorvaldsJul 30, 2008
  19. Linus TorvaldsJul 30, 2008
  20. Junio C HamanoJul 30, 2008
  21. Junio C HamanoJul 31, 2008
  22. Linus TorvaldsJul 31, 2008
  23. revision traversal: show full history with merge simplificationJunio C Hamano, Jul 31, 2008
  24. Junio C HamanoJul 31, 2008
  25. Linus TorvaldsJul 31, 2008
  26. revision traversal: show full history with merge simplificationJunio C Hamano, Jul 31, 2008
  27. Linus TorvaldsJul 31, 2008
  28. Junio C HamanoJul 31, 2008
  29. Junio C HamanoAug 1, 2008
  30. Linus TorvaldsAug 1, 2008
  31. Junio C HamanoAug 1, 2008
  32. Jakub NarebskiJul 30, 2008
  33. Linus TorvaldsJul 29, 2008
  34. Linus TorvaldsJul 29, 2008
  35. Roman ZippelJul 29, 2008
  36. David KastrupJul 29, 2008
  37. Linus TorvaldsJul 29, 2008
  38. Roman ZippelJul 30, 2008
  39. Kevin BallardJul 30, 2008
  40. Linus TorvaldsJul 30, 2008
  41. Jeff KingJul 29, 2008
  42. Roman ZippelJul 29, 2008
  43. Olivier GalibertJul 29, 2008
  44. Jeff KingJul 29, 2008
  45. Linus TorvaldsJul 29, 2008
  46. Roman ZippelJul 30, 2008
  47. Linus TorvaldsJul 30, 2008
  48. Jeff KingJul 30, 2008
  49. Linus TorvaldsJul 30, 2008
  50. Roman ZippelJul 30, 2008
  51. Kevin BallardJul 30, 2008
  52. Linus TorvaldsJul 30, 2008
  53. Linus TorvaldsJul 30, 2008
  54. Jeff KingJul 30, 2008
  55. Martin LanghoffJul 27, 2008
  56. Roman ZippelJul 28, 2008
  57. Alex RiesenJul 21, 2008
  58. Linus TorvaldsJul 21, 2008

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.