Re: [PATCH v2] [GSoC] submodule--helper: introduce add-clone subcommand

[Atharva Raykar <raykar.ath@gmail.com>]

On 04-Jun-2021, at 13:51, Christian Couder <christian.couder@gmail.com> wrote:
> 
> On Wed, Jun 2, 2021 at 3:13 PM Atharva Raykar <raykar.ath@gmail.com> wrote:
> 
>> +static void show_fetch_remotes(FILE *output, const char *sm_name, const char *git_dir_path)
>> +{
>> +       struct child_process cp_remote = CHILD_PROCESS_INIT;
>> +       struct strbuf sb_remote_out = STRBUF_INIT;
>> +
>> +       cp_remote.git_cmd = 1;
>> +       strvec_pushf(&cp_remote.env_array,
>> +                    "GIT_DIR=%s", git_dir_path);
>> +       strvec_push(&cp_remote.env_array, "GIT_WORK_TREE=.");
>> +       strvec_pushl(&cp_remote.args, "remote", "-v", NULL);
>> +       if (!capture_command(&cp_remote, &sb_remote_out, 0)) {
>> +               char *line;
>> +               char *begin = sb_remote_out.buf;
>> +               char *end = sb_remote_out.buf + sb_remote_out.len;
>> +               while (begin != end && (line = get_next_line(begin, end))) {
>> +                       char *name, *url, *tail;
>> +                       name = parse_token(&begin, line);
>> +                       url = parse_token(&begin, line);
>> +                       tail = parse_token(&begin, line);
> 
> Sorry for not replying to your earlier message, but I think it's a bit
> better to save a line with:
> 
>                       char *name = parse_token(&begin, line);
>                       char *url = parse_token(&begin, line);
>                       char *tail = parse_token(&begin, line);

Alright.

>> +                       if (!memcmp(tail, "(fetch)", 7))
>> +                               fprintf(output, "  %s\t%s\n", name, url);
>> +                       free(url);
>> +                       free(name);
>> +                       free(tail);
>> +               }
>> +       }
>> +
>> +       strbuf_release(&sb_remote_out);
>> +}
>> +
>> +static int add_submodule(const struct add_data *info)
>> +{
>> +       char *submod_gitdir_path;
>> +       /* perhaps the path already exists and is already a git repo, else clone it */
>> +       if (is_directory(info->sm_path)) {
>> +               printf("sm_path=%s\n", info->sm_path);
> 
> I don't see which shell code the above printf(...) instruction is
> replacing. That's why I asked if it's some debugging leftover.

Oh, my bad. It is a leftover debugging statement. Please excuse
my temporary blindness to it (:

> [...]
> 
>> +               if (info->dissociate)
>> +                       strvec_push(&clone_args, "--dissociate");
>> +               if (info->depth >= 0)
>> +                       strvec_pushf(&clone_args, "--depth=%d", info->depth);
> 
> It's ok if there is a blank line here.

OK. Makes sense.

>> +               if (module_clone(clone_args.nr, clone_args.v, info->prefix)) {
>> +                       strvec_clear(&clone_args);
>> +                       return -1;
>> +               }
>> +               strvec_clear(&clone_args);
> 
>> +static int add_clone(int argc, const char **argv, const char *prefix)
>> +{
>> +       const char *branch = NULL, *sm_path = NULL;
>> +       const char *wt_prefix = NULL, *realrepo = NULL;
>> +       const char *reference = NULL, *sm_name = NULL;
>> +       int force = 0, quiet = 0, dissociate = 0, depth = -1, progress = 0;
>> +       struct add_data info = ADD_DATA_INIT;
> 
> Maybe: s/info/add_data/

'info' was the local convention for naming similar structures that
held the flag values (like summary_cb, module_cb, deinit_cb etc).

The exception to the above is 'struct submodule_update_clone', which
was named as 'suc'. It did not follow the *_cb naming convention,
presumably because it was not used as a parameter passed to any
*_cb() function.

Since 'struct add_data' is more similar to the latter (as it is not
used in any callback function) I guess it would be okay to name it
differently and more descriptively as 'add_data'?

> Also it seems that in many cases it's a bit wasteful to use new
> variables for option parsing and then to copy them into the add_data
> struct when the field of the add_data struct could be used directly
> for option parsing...
> 
>> +       struct option options[] = {
>> +               OPT_STRING('b', "branch", &branch,
> 
> ...for example, here maybe `&add_data.branch` could be used instead of
> `&branch`...

I thought of this too, but decided to stick to the surrounding
convention, where a new variable is used and then assigned to the
struct.

I had a looked at the file again, and turns out...

	OPT_STRING_LIST(0, "reference", &suc.references, N_("repo"),
		   N_("reference repository")),
	OPT_BOOL(0, "dissociate", &suc.dissociate,
		   N_("use --reference only while cloning")),
	OPT_STRING(0, "depth", &suc.depth, "<depth>",
		   N_("create a shallow clone truncated to the "
		      "specified number of revisions")),

... update_clone() is the exception again.

So there is precedent, and I'd rather follow what you suggested,
because that looks much better to me, and saves a lot of redundant
code.

>> +                          N_("branch"),
>> +                          N_("branch of repository to checkout on cloning")),
> 
> [...]
> 
>> +       info.branch = branch;
> 
> ...so that the above line would not be needed.

Yes, although I might still need to use an extra variable for
booleans, like 'progress' or 'dissociate', because of the need
to use !! to make it either 1 or 0. I am not too familiar with
why doing that would be important in this context, but since
this is the convention, I'll keep it intact.