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

Re: [PATCH 1/2] difftool: use "struct strvec" API in run_{dir,file}_diff()

From
Jeff King <peff@peff.net>
Date
Sep 12, 2021, 22:39 UTC
Message-ID
<YT6BnnXeAWn8BycF@coredump.intra.peff.net>
In-Reply-To
<patch-1.2-e7481eb0c0c-20210911T182009Z-avarab@gmail.com>
On Sat, Sep 11, 2021 at 08:21:11PM +0200, Ævar Arnfjörð Bjarmason wrote:
> The underlying run_command() API can take either the "struct strvec
> args", or a "const char **argv". Let's move to the former to use the
> more "native" version of run_command() in both of these functions.

It sounds like we're moving to use child.args (the strvec interface) instead of child.argv (the const char one). Which I support; I'd like to eventually get rid of the argv interface entirely because it has memory-ownership semantics that are easy to get wrong.

But this...
Show 10 quoted lines
> @@ -393,10 +393,8 @@ static int run_dir_diff(const char *extcmd, int symlinks, const char *prefix,
>  	child.clean_on_exit = 1;
>  	child.dir = prefix;
>  	child.out = -1;
> -	strvec_pushl(&child.args, "diff", "--raw", "--no-abbrev", "-z",
> -		     NULL);
> -	for (i = 0; i < argc; i++)
> -		strvec_push(&child.args, argv[i]);
> +	child.argv = args->v;
> +
...is going in the opposite direction.
I'd much rather see us continue to use child.args here, like:
  strvec_pushv(&child.args, args->v);

Though really I do think passing the strvec into run_dir_diff() is questionable in the first place. The caller depends on us to free the memory in the strvec for them, which is...subtle.

It does let you immediately return here:
Show 5 quoted lines
>  	if (dir_diff)
> -		return run_dir_diff(extcmd, symlinks, prefix, argc, argv);
> -	return run_file_diff(prompt, prefix, argc, argv);
> +		return run_dir_diff(extcmd, symlinks, prefix, &args);
> +	return run_file_diff(prompt, prefix, &args);
without doing any cleanup. But I think just:
  if (dir_diff)
	ret = run_dir_diff(...);
  else
	ret = run_file_diff(...);
  strvec_clear(&args);
  return ret;
would be a lot more obvious.

I almost suggested that this could be done even simpler by having the caller pass in one of two pre-made argv lists (in addition to what's in the original argv). But the _actual_ benefit you want here is making it easy to construct those lists in the shared code of the caller. Your commit message didn't at all make that clear (so to me it looked like your "cleanup" was actively making things worse).

It only becomes apparent with the second patch. I would have found it much easier to understand with something like the patch below. And then a further patch to use strvec_pushv instead of manually looping (even getting rid of the argc parameters entirely!), and one to convert run_file_diff() to use a struct child_process (which fixes its memory leak).

-- >8 -- difftool: prepare "diff" cmdline in cmd_difftool()

We call into either run_dir_diff() or run_file_diff(), each of which sets up a child argv starting with "diff" and some hard-coded options (depending on which mode we're using). Let's extract that logic into the caller, which will make it easier to modify the options for cases which affect both functions.

Signed-off-by: Jeff King <peff@peff.net>
---
 builtin/difftool.c | 13 ++++++++-----
 1 file changed, 8 insertions(+), 5 deletions(-)
diff --git a/builtin/difftool.c b/builtin/difftool.c
index 6a9242a803..91a8e51b0c 100644
--- a/builtin/difftool.c
+++ b/builtin/difftool.c
@@ -393,8 +393,6 @@ static int run_dir_diff(const char *extcmd, int symlinks, const char *prefix,
 	child.clean_on_exit = 1;
 	child.dir = prefix;
 	child.out = -1;
-	strvec_pushl(&child.args, "diff", "--raw", "--no-abbrev", "-z",
-		     NULL);
 	for (i = 0; i < argc; i++)
 		strvec_push(&child.args, argv[i]);
 	if (start_command(&child))
@@ -683,7 +681,6 @@ static int run_file_diff(int prompt, const char *prefix,
 		env[2] = "GIT_DIFFTOOL_NO_PROMPT=true";
 
 
-	strvec_push(&args, "diff");
 	for (i = 0; i < argc; i++)
 		strvec_push(&args, argv[i]);
 	return run_command_v_opt_cd_env(args.v, RUN_GIT_CMD, prefix, env);
@@ -719,6 +716,7 @@ int cmd_difftool(int argc, const char **argv, const char *prefix)
 		OPT_ARGUMENT("no-index", &no_index, N_("passed to `diff`")),
 		OPT_END()
 	};
+	struct strvec args = STRVEC_INIT;
 
 	git_config(difftool_config, NULL);
 	symlinks = has_symlinks;
@@ -768,7 +766,12 @@ int cmd_difftool(int argc, const char **argv, const char *prefix)
 	 * will invoke a separate instance of 'git-difftool--helper' for
 	 * each file that changed.
 	 */
+	strvec_push(&args, "diff");
+	if (dir_diff)
+		strvec_pushl(&args, "--raw", "--no-abbrev", "-z", NULL);
+	strvec_pushv(&args, argv);
+
 	if (dir_diff)
-		return run_dir_diff(extcmd, symlinks, prefix, argc, argv);
-	return run_file_diff(prompt, prefix, argc, argv);
+		return run_dir_diff(extcmd, symlinks, prefix, args.nr, args.v);
+	return run_file_diff(prompt, prefix, args.nr, args.v);
 }
-- 
2.33.0.811.g40f7f3a89c
Previous: Ævar Arnfjörð BjarmasonNext: Jeff King
Message 3 of 17 in “parse-options.c: remove OPT_ARGUMENT”
  1. 0/2 parse-options.c: remove OPT_ARGUMENTÆvar Arnfjörð Bjarmason, Sep 11, 2021
  2. 1/2 difftool: use "struct strvec" API in run_{dir,file}_diff()Ævar Arnfjörð Bjarmason, Sep 11, 2021
  3. Jeff KingSep 12, 2021
  4. Jeff KingSep 12, 2021
  5. Junio C HamanoSep 13, 2021
  6. 2/2 parse-options API: remove OPTION_ARGUMENT featureÆvar Arnfjörð Bjarmason, Sep 11, 2021
  7. Jeff KingSep 12, 2021
  8. 0/4 difftool refactoring + remove OPT_ARGUMENT() macroÆvar Arnfjörð Bjarmason, Sep 13, 2021
  9. 2/4 difftool: prepare "diff" cmdline in cmd_difftool()Ævar Arnfjörð Bjarmason, Sep 13, 2021
  10. 1/4 difftool: prepare "struct child_process" in cmd_difftool()Ævar Arnfjörð Bjarmason, Sep 13, 2021
  11. 3/4 difftool: use run_command() API in run_file_diff()Ævar Arnfjörð Bjarmason, Sep 13, 2021
  12. Jeff KingSep 13, 2021
  13. Ævar Arnfjörð BjarmasonSep 13, 2021
  14. Jeff KingSep 13, 2021
  15. 4/4 parse-options API: remove OPTION_ARGUMENT featureÆvar Arnfjörð Bjarmason, Sep 13, 2021
  16. Junio C HamanoSep 13, 2021
  17. Jeff KingSep 13, 2021

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.