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

Re: [PATCH] am: terminate state files with a newline

From
Junio C Hamano <gitster@pobox.com>
Date
Aug 23, 2015, 19:05 UTC
Message-ID
<xmqqy4h16d1f.fsf@gitster.dls.corp.google.com>
In-Reply-To
<20150823055053.GA15849@yoshi.chippynet.com>
Paul Tan <pyokagan@gmail.com> writes:
> Did we ever explictly allow external programs to poke around the
> contents of the .git/rebase-apply directory?

We tell users to take a peek into it when "am" fails, don't we, by naming $GIT_DIR/rebase-apply/patch?

> -	write_file(am_path(state, "threeway"), 1, state->threeway ? "t" : "f");
> +	write_file(am_path(state, "threeway"), 1, "%s\n", state->threeway ? "t" : "f");

Stepping back a bit, after realizing that "write_file()" is a short-hand for "I have all information necessary to produce the full contents of a file, now go ahead and create and write that and close", I have to wonder what caller even wants to create a file with an incomplete line at the end.

All callers outside builtin/am.c except one caller uses it to produce a single line file. The oddball is "git branch" that uses it to prepare a temporary file used to edit branch description.

builtin/branch.c:	if (write_file(git_path(edit_description), 0, "%s", buf.buf)) {

The payload it prepares in buf.buf ends with a canned comment that ends with LF. So in that sense it is not even an oddball.

The above analysis makes me wonder if this is a simpler and more future proof approach.

Or did I miss any caller or a reasonable potential future use case that wants to create a binary file or a text file that ends with an incomplete line?

 wrapper.c | 10 ++++++++++
 1 file changed, 10 insertions(+)
diff --git a/wrapper.c b/wrapper.c
index e451463..7a92298 100644
--- a/wrapper.c
+++ b/wrapper.c
@@ -621,6 +621,13 @@ char *xgetcwd(void)
 	return strbuf_detach(&sb, NULL);
 }
 
+/*
+ * Create a TEXT file by specifying its full contents via fmt and the
+ * remainder of args that are used like "printf".  A terminating LF is
+ * added at the end of the file if it is missing (it is simpler for
+ * the callers because the function is often used to create a
+ * single-liner file).
+ */
 int write_file(const char *path, int fatal, const char *fmt, ...)
 {
 	struct strbuf sb = STRBUF_INIT;
@@ -634,6 +641,9 @@ int write_file(const char *path, int fatal, const char *fmt, ...)
 	va_start(params, fmt);
 	strbuf_vaddf(&sb, fmt, params);
 	va_end(params);
+	if (sb.len)
+		strbuf_complete_line(&sb);
+
 	if (write_in_full(fd, sb.buf, sb.len) != sb.len) {
 		int err = errno;
 		close(fd);
Previous: SZEDER GáborNext: Jeff King
Message 5 of 36 in “Minor builtin 'git am' side-effect”
  1. SZEDER GáborAug 20, 2015
  2. Junio C HamanoAug 20, 2015
  3. am: terminate state files with a newlinePaul Tan, Aug 23, 2015
  4. SZEDER GáborAug 23, 2015
  5. Junio C HamanoAug 23, 2015
  6. Jeff KingAug 24, 2015
  7. Junio C HamanoAug 24, 2015
  8. Jeff KingAug 24, 2015
  9. 0/5 "am" state file fix with write_file() clean-upJunio C Hamano, Aug 24, 2015
  10. 1/5 builtin/am: introduce write_state_*() helper functionsJunio C Hamano, Aug 24, 2015
  11. 2/5 builtin/am: make sure state files are textJunio C Hamano, Aug 24, 2015
  12. 3/5 write_file(): introduce an explicit WRITE_FILE_GENTLY requestJunio C Hamano, Aug 24, 2015
  13. Junio C HamanoAug 24, 2015
  14. Duy NguyenAug 25, 2015
  15. setup: update the right file in multiple checkoutsNguyễn Thái Ngọc Duy, Aug 25, 2015
  16. Junio C HamanoAug 25, 2015
  17. Duy NguyenAug 31, 2015
  18. 4/5 write_file(): do not leave incomplete line at the endJunio C Hamano, Aug 24, 2015
  19. 5/5 write_file(): clean up transitional mess of flag words and terminating LFJunio C Hamano, Aug 24, 2015
  20. Jeff KingAug 24, 2015
  21. Junio C HamanoAug 24, 2015
  22. Jeff KingAug 24, 2015
  23. Junio C HamanoAug 24, 2015
  24. 0/6 "am" state file fix with write_file() clean-upJunio C Hamano, Aug 24, 2015
  25. 1/6 builtin/am: introduce write_state_*() helper functionsJunio C Hamano, Aug 24, 2015
  26. 2/6 builtin/am: make sure state files are textJunio C Hamano, Aug 24, 2015
  27. Jeff KingAug 24, 2015
  28. Junio C HamanoAug 25, 2015
  29. Jeff KingAug 25, 2015
  30. Junio C HamanoAug 25, 2015
  31. 3/6 write_file(): drop "fatal" parameterJunio C Hamano, Aug 24, 2015
  32. 4/6 write_file_v(): do not leave incomplete line at the endJunio C Hamano, Aug 24, 2015
  33. 5/6 write_file(): drop caller-supplied LF from calls to create a one-liner fileJunio C Hamano, Aug 24, 2015
  34. 6/6 write_file(): drop caller-supplied LF from multi-line fileJunio C Hamano, Aug 24, 2015
  35. Jeff KingAug 25, 2015
  36. brian m. carlsonAug 24, 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.