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

Re: [PATCH v2 09/13] Honour the refspec when updating refs after import

From
Daniel Barkalow <barkalow@iabervon.org>
Date
Nov 4, 2009, 21:20 UTC
Message-ID
<alpine.LNX.2.00.0911041601170.14365@iabervon.org>
In-Reply-To
<1257364098-1685-10-git-send-email-srabbelier@gmail.com>
On Wed, 4 Nov 2009, Sverre Rabbelier wrote:
Show 25 quoted lines
> This way the helper can write to 'refs/remotes/origin/*' instead of
> writing to 'refs/heads/*'.
> 
> Signed-off-by: Sverre Rabbelier <srabbelier@gmail.com>
> ---
> 
>   Daniel, does this look good to you?
> 
>   Let's assume the following:
>   * list: only HEAD is a symref (to refs/heads/<branch>), all other
>           refs are '? refs/heads/<branch>'.
>   * import: the helper creates branches under refs/remotes/<alias>/*
>             (since the refspec code is not yet in that would allow
>              it to create them under refs/<vcs>/<alias>/*)
>
>   Now when updating the refs we do the following:
>   transport-helper.c:145 "read_ref(posn->name, posn->old_sha1);"
> 
>   The value of posn->name looks like "refs/heads/<branch>". So what
>   does this lookup do? It tries to lookup the new value of the ref
>   _where it will be created_, this fails of course, since the ref
>   currently only exists as "refs/remotes/origin/<branch>". So, we
>   should instead be doing a lookup using posn->peer_ref->name, and
>   not do a lookup at all if it posn->peer_ref is not set (which
>   should not occur as we are passed only wanted peer refs).

That's not true for "git pull <url> <branch>"; we do want the remote ref, but it doesn't have a local peer. I think going straight to the refspec command is the right answer.

	-Daniel
*This .sig left intentionally blank*
Previous: Sverre RabbelierNext: Sverre Rabbelier
Message 11 of 36 in “Reroll of the remote-vcs-helper series”
  1. 00/13 Reroll of the remote-vcs-helper seriesSverre Rabbelier, Nov 4, 2009
  2. 01/13 Fix memory leak in helper method for disconnectSverre Rabbelier, Nov 4, 2009
  3. 02/13 Allow programs to not depend on remotes having urlsSverre Rabbelier, Nov 4, 2009
  4. 03/13 Use a function to determine whether a remote is validSverre Rabbelier, Nov 4, 2009
  5. 04/13 Allow fetch to modify refsSverre Rabbelier, Nov 4, 2009
  6. 05/13 Add a config option for remotes to specify a foreign vcsSverre Rabbelier, Nov 4, 2009
  7. 06/13 Allow specifying the remote helper in the urlSverre Rabbelier, Nov 4, 2009
  8. 07/13 Add support for "import" helper commandSverre Rabbelier, Nov 4, 2009
  9. 08/13 Write local refs written by the "import" helper command only onceSverre Rabbelier, Nov 4, 2009
  10. 09/13 Honour the refspec when updating refs after importSverre Rabbelier, Nov 4, 2009
  11. Daniel BarkalowNov 4, 2009
  12. Sverre RabbelierNov 4, 2009
  13. Daniel BarkalowNov 4, 2009
  14. Sverre RabbelierNov 5, 2009
  15. Daniel BarkalowNov 5, 2009
  16. Daniel BarkalowNov 5, 2009
  17. Sverre RabbelierNov 5, 2009
  18. Daniel BarkalowNov 6, 2009
  19. Sverre RabbelierNov 6, 2009
  20. Daniel BarkalowNov 6, 2009
  21. Junio C HamanoNov 6, 2009
  22. 10/13 Allow helpers to report in "list" command that the ref is unchangedSverre Rabbelier, Nov 4, 2009
  23. 11/13 Allow helpers to request the path to the .git directorySverre Rabbelier, Nov 4, 2009
  24. Daniel BarkalowNov 4, 2009
  25. Sverre RabbelierNov 4, 2009
  26. Daniel BarkalowNov 4, 2009
  27. Sverre RabbelierNov 4, 2009
  28. Daniel BarkalowNov 4, 2009
  29. Sverre RabbelierNov 4, 2009
  30. Daniel BarkalowNov 4, 2009
  31. Junio C HamanoNov 5, 2009
  32. Sverre RabbelierNov 5, 2009
  33. 12/13 Basic build infrastructure for Python scriptsSverre Rabbelier, Nov 4, 2009
  34. 13/13 Add Python support library for remote helpersSverre Rabbelier, Nov 4, 2009
  35. Johan HerlandNov 5, 2009
  36. Sverre RabbelierNov 6, 2009

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.