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

Re: [PATCH v2 3/4] status: show more info than "currently not on any branch"

From
Junio C Hamano <gitster@pobox.com>
Date
Mar 6, 2013, 19:16 UTC
Message-ID
<7vzjygpcd5.fsf@alter.siamese.dyndns.org>
In-Reply-To
<1362572482-20570-4-git-send-email-pclouds@gmail.com>
Nguyễn Thái Ngọc Duy  <pclouds@gmail.com> writes:
Show 12 quoted lines
> +static void wt_status_get_detached_from(struct wt_status_state *state)
> +{
> +	struct grab_1st_switch_cbdata cb;
> +	struct commit *commit;
> +	unsigned char sha1[20];
> +	char *ref = NULL;
> +
> +	strbuf_init(&cb.buf, 0);
> +	if (for_each_recent_reflog_ent("HEAD", grab_1st_switch, 4096, &cb))
> +		for_each_reflog_ent("HEAD", grab_1st_switch, &cb);
> +	if (!cb.buf.len)
> +		return;

Is this correct? What if the recent entries (i.e. the tail 4k) of the HEAD reflog did not have *any* checkout? Your callback never returns non-zero, so as long as the HEAD reflog is sufficiently long, for_each_recent_reflog_ent() will return 0 to signal success, and you do not dig deeper by retrying the full reflog for HEAD, missing the checkout that exists before the final 4k, no?

It should be more like this, I would think:
	for_each_recent_reflog_ent();
        if (!found)
        	for_each_reflog_ent();
	if (!found)
        	return;

Using cb.buf.len as the "found" marker may be correct, but I found it a bit subtle to my taste, without explanation. Adding an explicit bit to "struct grab_1st_switch_cbdata" would be cleaner and more resistant to future changes, I think.

> +
> +	if (dwim_ref(cb.buf.buf, cb.buf.len, sha1, &ref) == 1 &&
> +	    (commit = lookup_commit_reference_gently(sha1, 1)) != NULL &&
> +	    !hashcmp(cb.nsha1, commit->object.sha1)) {

That feels unnecessarily expensive. Why not hashcmp before checking the type of the object to reject the case where the ref moved since the last checkout early?

For that matter, does this even need to check the type of the object that currently sits at the ref? Isn't it sufficient to reject this case by seeing if sha1 is the same as cb.nsha1?

Previous: Nguyễn Thái Ngọc DuyNext: Duy Nguyen
Message 19 of 38 in “nd/branch-show-rebase-bisect-state updates”
  1. 0/5 nd/branch-show-rebase-bisect-state updatesNguyễn Thái Ngọc Duy, Mar 3, 2013
  2. 1/5 checkout: record full target ref in reflogNguyễn Thái Ngọc Duy, Mar 3, 2013
  3. 2/5 wt-status: split wt_status_state parsing function outNguyễn Thái Ngọc Duy, Mar 3, 2013
  4. 3/5 wt-status: move wt_status_get_state() out to wt_status_print()Nguyễn Thái Ngọc Duy, Mar 3, 2013
  5. 4/5 status: show the ref that is checked out, even if it's detachedNguyễn Thái Ngọc Duy, Mar 3, 2013
  6. Junio C HamanoMar 3, 2013
  7. Duy NguyenMar 4, 2013
  8. Junio C HamanoMar 4, 2013
  9. Duy NguyenMar 5, 2013
  10. Matthieu MoyMar 5, 2013
  11. 5/5 branch: show more information when HEAD is detachedNguyễn Thái Ngọc Duy, Mar 3, 2013
  12. Junio C HamanoMar 3, 2013
  13. 0/4 nd/branch-show-rebase-bisect-state updatesNguyễn Thái Ngọc Duy, Mar 6, 2013
  14. 1/4 wt-status: split wt_status_state parsing function outNguyễn Thái Ngọc Duy, Mar 6, 2013
  15. Junio C HamanoMar 6, 2013
  16. Junio C HamanoMar 6, 2013
  17. 2/4 wt-status: move wt_status_get_state() out to wt_status_print()Nguyễn Thái Ngọc Duy, Mar 6, 2013
  18. 3/4 status: show more info than "currently not on any branch"Nguyễn Thái Ngọc Duy, Mar 6, 2013
  19. Junio C HamanoMar 6, 2013
  20. Duy NguyenMar 8, 2013
  21. Junio C HamanoMar 8, 2013
  22. 4/4 branch: show more information when HEAD is detachedNguyễn Thái Ngọc Duy, Mar 6, 2013
  23. 0/4 nd/branch-show-rebase-bisect-state updatesNguyễn Thái Ngọc Duy, Mar 6, 2013
  24. 4/4 branch: show more information when HEAD is detachedNguyễn Thái Ngọc Duy, Mar 6, 2013
  25. 0/5 nd/branch-show-rebase-bisect-state updatesNguyễn Thái Ngọc Duy, Mar 13, 2013
  26. 1/5 wt-status: move strbuf into read_and_strip_branch()Nguyễn Thái Ngọc Duy, Mar 13, 2013
  27. Junio C HamanoMar 13, 2013
  28. 1/5 wt-status: move strbuf into read_and_strip_branch()Nguyễn Thái Ngọc Duy, Mar 16, 2013
  29. 2/5 wt-status: split wt_status_state parsing function outNguyễn Thái Ngọc Duy, Mar 13, 2013
  30. 3/5 wt-status: move wt_status_get_state() out to wt_status_print()Nguyễn Thái Ngọc Duy, Mar 13, 2013
  31. 4/5 status: show more info than "currently not on any branch"Nguyễn Thái Ngọc Duy, Mar 13, 2013
  32. Junio C HamanoMar 13, 2013
  33. 5/5 branch: show more information when HEAD is detachedNguyễn Thái Ngọc Duy, Mar 13, 2013
  34. Junio C HamanoMar 19, 2013
  35. Duy NguyenMar 20, 2013
  36. Junio C HamanoMar 20, 2013
  37. status, branch: fix the misleading "bisecting" messageNguyễn Thái Ngọc Duy, Mar 23, 2013
  38. Junio C HamanoMar 24, 2013

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.