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

[PATCH v2 03/13] rerere: lift PATH_MAX limitation

From
Junio C Hamano <gitster@pobox.com>
Date
Jul 1, 2015, 06:04 UTC
Message-ID
<1435730699-9124-4-git-send-email-gitster@pobox.com>
In-Reply-To
<1435730699-9124-1-git-send-email-gitster@pobox.com>

The MERGE_RR file records a collection of NUL-terminated entries, each of which consists of

 - a hash that identifies the conflict
 - a HT
 - the pathname

We used to read this piece-by-piece, and worse yet, read the pathname part a byte at a time into a fixed buffer of size PATH_MAX.

Instead, read a whole entry using strbuf_getwholeline() and parse out the fields. This way, we issue fewer read(2) calls and more importantly we do not have to limit the pathname to PATH_MAX.

Signed-off-by: Junio C Hamano <gitster@pobox.com>
---
 rerere.c | 35 +++++++++++++++--------------------
 1 file changed, 15 insertions(+), 20 deletions(-)
diff --git a/rerere.c b/rerere.c
index 3b9104d..e854985 100644
--- a/rerere.c
+++ b/rerere.c
@@ -35,32 +35,27 @@ static int has_rerere_resolution(const char *hex)
 
 static void read_rr(struct string_list *rr)
 {
-	unsigned char sha1[20];
-	char buf[PATH_MAX];
+	struct strbuf buf = STRBUF_INIT;
 	FILE *in = fopen(merge_rr_path, "r");
+
 	if (!in)
 		return;
-	while (fread(buf, 40, 1, in) == 1) {
-		int i;
-		char *name;
-		if (get_sha1_hex(buf, sha1))
+	while (!strbuf_getwholeline(&buf, in, '\0')) {
+		char *path;
+		unsigned char sha1[20];
+
+		/* There has to be the hash, tab, path and then NUL */
+		if (buf.len < 42 || get_sha1_hex(buf.buf, sha1))
 			die("corrupt MERGE_RR");
-		buf[40] = '\0';
-		name = xstrdup(buf);
-		if (fgetc(in) != '\t')
+
+		if (buf.buf[40] != '\t')
 			die("corrupt MERGE_RR");
-		for (i = 0; i < sizeof(buf); i++) {
-			int c = fgetc(in);
-			if (c < 0)
-				die("corrupt MERGE_RR");
-			buf[i] = c;
-			if (c == 0)
-				 break;
-		}
-		if (i == sizeof(buf))
-			die("filename too long");
-		string_list_insert(rr, buf)->util = name;
+		buf.buf[40] = '\0';
+		path = buf.buf + 41;
+
+		string_list_insert(rr, path)->util = xstrdup(buf.buf);
 	}
+	strbuf_release(&buf);
 	fclose(in);
 }
 
-- 
2.5.0-rc0-209-g5e1f148
Previous: Junio C HamanoNext: Junio C Hamano
Message 4 of 39 in “"rerere" minor clean-up”
  1. 00/13 "rerere" minor clean-upJunio C Hamano, Jul 1, 2015
  2. 01/13 rerere: fix an off-by-one non-bugJunio C Hamano, Jul 1, 2015
  3. 02/13 rerere: plug conflict ID leaksJunio C Hamano, Jul 1, 2015
  4. 03/13 rerere: lift PATH_MAX limitationJunio C Hamano, Jul 1, 2015
  5. 04/13 rerere: write out each record of MERGE_RR in one goJunio C Hamano, Jul 1, 2015
  6. 05/13 rerere: report autoupdated paths only after actually updating themJunio C Hamano, Jul 1, 2015
  7. 06/13 rerere: drop want_sp parameter from is_cmarker()Junio C Hamano, Jul 1, 2015
  8. 07/13 rerere: stop looping unnecessarilyJunio C Hamano, Jul 1, 2015
  9. 08/13 rerere: explain the rerere I/O abstractionJunio C Hamano, Jul 1, 2015
  10. 09/13 rerere: explain MERGE_RR management helpersJunio C Hamano, Jul 1, 2015
  11. 10/13 rerere: explain the primary codepathJunio C Hamano, Jul 1, 2015
  12. 11/13 rerere: explain "rerere forget" codepathJunio C Hamano, Jul 1, 2015
  13. 12/13 rerere: explain the remainderJunio C Hamano, Jul 1, 2015
  14. 13/13 rerere: refactor "replay" part of do_plain_rerere()Junio C Hamano, Jul 1, 2015
  15. 00/18 "rerere" preparatory clean-upJunio C Hamano, Jul 17, 2015
  16. 01/18 rerere: fix an off-by-one non-bugJunio C Hamano, Jul 17, 2015
  17. Junio C HamanoJul 24, 2015
  18. 02/18 rerere: plug conflict ID leaksJunio C Hamano, Jul 17, 2015
  19. 03/18 rerere: lift PATH_MAX limitationJunio C Hamano, Jul 17, 2015
  20. 04/18 rerere: write out each record of MERGE_RR in one goJunio C Hamano, Jul 17, 2015
  21. 05/18 rerere: report autoupdated paths only after actually updating themJunio C Hamano, Jul 17, 2015
  22. 06/18 rerere: drop want_sp parameter from is_cmarker()Junio C Hamano, Jul 17, 2015
  23. Philip OakleyJul 18, 2015
  24. Eric SunshineJul 18, 2015
  25. 07/18 rerere: stop looping unnecessarilyJunio C Hamano, Jul 17, 2015
  26. Junio C HamanoJul 24, 2015
  27. 08/18 rerere: explain the rerere I/O abstractionJunio C Hamano, Jul 17, 2015
  28. Junio C HamanoJul 24, 2015
  29. 09/18 rerere: explain MERGE_RR management helpersJunio C Hamano, Jul 17, 2015
  30. 10/18 rerere: explain the primary codepathJunio C Hamano, Jul 17, 2015
  31. 11/18 rerere: explain "rerere forget" codepathJunio C Hamano, Jul 17, 2015
  32. 12/18 rerere: explain the remainderJunio C Hamano, Jul 17, 2015
  33. 13/18 rerere: refactor "replay" part of do_plain_rerere()Junio C Hamano, Jul 17, 2015
  34. 14/18 rerere: further de-dent do_plain_rerere()Junio C Hamano, Jul 17, 2015
  35. 15/18 rerere: further clarify do_rerere_one_path()Junio C Hamano, Jul 17, 2015
  36. 16/18 rerere: call conflict-ids IDsJunio C Hamano, Jul 17, 2015
  37. 17/18 rerere: use "struct rerere_id" instead of "char *" for conflict IDJunio C Hamano, Jul 17, 2015
  38. Eric SunshineJul 18, 2015
  39. 18/18 rerere: un-nest merge() furtherJunio C Hamano, Jul 17, 2015

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.