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

Re: [PATCH 06/10] commit-graph: move remaining write parameters into options

From
Patrick Steinhardt <ps@pks.im>
Date
Nov 5, 2026, 13:20 UTC
Message-ID
<auyCsmn6Hx_uc6zq@pks.im>
In-Reply-To
<CAL71e4N4BvO=eMfO0wnJ9=gGbfTRz5ayWFZULoN1dGSF7uCZAA@mail.gmail.com>
On Wed, Nov 04, 2026 at 06:49:50PM +0100, Kristofer Karlsson wrote:
Show 29 quoted lines
> On Wed, 4 Nov 2026 at 17:15, Patrick Steinhardt <ps@pks.im> wrote:
> >
> >                 if (graph_write_mode != GRAPH_WRITE_TIPS)
> >                         write_opts.flags |= COMMIT_GRAPH_WRITE_REACHABLE;
> > +               else
> > +                       write_opts.commits = &updated_tips;
> 
> Ah, this touches on my previous patch cef2d8c4cb
> (fetch: write commit-graph using updated refs only, 2026-10-07).
> 
> My only nit would be that the if/else also covers the skip case,
> so one of the branches always fires for it.  A switch would make
> each mode set exactly what it needs:
> 
>   switch (graph_write_mode) {
>   case GRAPH_WRITE_SKIP:
>       break;
>   case GRAPH_WRITE_REACHABLE:
>       write_opts.flags |= COMMIT_GRAPH_WRITE_REACHABLE;
>       break;
>   case GRAPH_WRITE_TIPS:
>       write_opts.commits = &updated_tips;
>       break;
>   }
> 
> It doesn't matter for correctness since the skip case never calls
> the writer, but it looks easier to reason about to me.
> 
> But that's just a preference, I don't feel strongly about it.

I was wondering whether we want to maybe introduce a tagged union where it becomes implicitly clear that some of the options only apply to a specific mode. Something like this:

diff --git a/commit-graph.h b/commit-graph.h
index 2156499a688..8d53f80f179 100644
--- a/commit-graph.h
+++ b/commit-graph.h
@@ -160,7 +160,6 @@ enum commit_graph_write_flags {
 	COMMIT_GRAPH_WRITE_SPLIT      = (1 << 2),
 	COMMIT_GRAPH_WRITE_BLOOM_FILTERS = (1 << 3),
 	COMMIT_GRAPH_NO_WRITE_BLOOM_FILTERS = (1 << 4),
-	COMMIT_GRAPH_WRITE_REACHABLE  = (1 << 5),
 };
 
 enum commit_graph_split_flags {
@@ -170,9 +169,18 @@ enum commit_graph_split_flags {
 };
 
 struct commit_graph_write_options {
+	enum commit_graph_write_mode {
+		COMMIT_GRAPH_WRITE_ALL,
+		COMMIT_GRAPH_WRITE_REACHABLE,
+		COMMIT_GRAPH_WRITE_PACKS_INDEXES,
+		COMMIT_GRAPH_WRITE_COMMITS,
+	} mode;
+	union {
+		const struct string_list *pack_indexes;
+		struct oidset *commits;
+	} mode_args;
+
 	enum commit_graph_write_flags flags;
-	const struct string_list *pack_indexes;
-	struct oidset *commits;
 	int size_multiple;
 	int max_commits;
 	timestamp_t expire_time;

I wasn't really sure whether that complexity is really worth it, so for
now I decided to not pursue it. I think I still tend to stick with that
decision for now, but it's maybe something to keep in mind for the
future?

Patrick
Previous: Kristofer KarlssonNext: Kristofer Karlsson
Message 12 of 22 in “commit-graph: several refactorings for writing commit graphs”
  1. 00/10 commit-graph: several refactorings for writing commit graphsPatrick Steinhardt, Nov 4, 2026
  2. 01/10 commit-graph: rename `struct commit_graph_opts`Patrick Steinhardt, Nov 4, 2026
  3. 02/10 commit-graph: pass writer flags via optionsPatrick Steinhardt, Nov 4, 2026
  4. 03/10 commit-graph: drop `NULL` checks for write optionsPatrick Steinhardt, Nov 4, 2026
  5. 04/10 commit-graph: drop redundant flags in write contextPatrick Steinhardt, Nov 4, 2026
  6. Junio C HamanoNov 4, 2026
  7. 05/10 commit-graph: merge the two commit graph writersPatrick Steinhardt, Nov 4, 2026
  8. Derrick StoleeNov 4, 2026
  9. 06/10 commit-graph: move remaining write parameters into optionsPatrick Steinhardt, Nov 4, 2026
  10. Derrick StoleeNov 4, 2026
  11. Kristofer KarlssonNov 4, 2026
  12. Patrick SteinhardtNov 5, 2026
  13. Kristofer KarlssonNov 5, 2026
  14. 07/10 commit-graph: fix mixup of write/verify flagsPatrick Steinhardt, Nov 4, 2026
  15. 08/10 builtin/commit-graph: set flags directlyPatrick Steinhardt, Nov 4, 2026
  16. 09/10 odb/source-files: verify commit-graphs as part of fsckPatrick Steinhardt, Nov 4, 2026
  17. 10/10 commit-graph: modernize function namesPatrick Steinhardt, Nov 4, 2026
  18. Derrick StoleeNov 4, 2026
  19. Patrick SteinhardtNov 5, 2026
  20. Patrick SteinhardtNov 5, 2026
  21. Derrick StoleeNov 5, 2026
  22. Patrick SteinhardtNov 5, 2026

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.