list mirror (unofficial, one of many)
 help / color / mirror / code / Atom feed
From: Christian Couder <>
To: Chinmoy Chakraborty <>
Cc: git <>, Kaartic Sivaraam <>
Subject: Re: [GSoC] Draft Proposal (Convert submodule to builtin)
Date: Sat, 10 Apr 2021 14:03:43 +0200	[thread overview]
Message-ID: <> (raw)
In-Reply-To: <>

On Thu, Apr 8, 2021 at 11:10 AM Chinmoy Chakraborty
<> wrote:
> This is the V2 of the draft proposal after applying the
> changes suggested by Christian Couder and Kaartik Sivaraam.

Please add all the reviewers in Cc. I added Kaartic.

> ##Git Contributions##
> [Microproject] Replace instances of `the_repository` with ‘r’. (Learning
> the ropes)
> Pull request:
> Mailing List:
> [column, range-diff] downcase option description
> Pull request:
> Mailing List:
> [Documentation] updated documentation for git commit --date
> Pull request:
> Mailing List:

Adding a status to let us know if each of your contributions have been
merged to seen, next or master, or even already released (in which
version?) would be nice. You could also add a GitHub or GitLab link to
the merge commit that merged your contribution.

> ##Project Outline##
> A few components of git, like ``
> are in the form of shell scripts. This causes
> problems in production code in multiple platforms
> like windows. The goal of this project is to
> convert the shell script version of ``
> to portable c code. The end goal would be


> to completely remove `` and rename
> `builtin/submodule--helper.c` to `builtin/submodule.c`
> so that the `git submodule` is fully implemented using C.
> =================================================
> ##Why is the project required?##
> "Issues with the portability of code"
> The submodule shell script uses shell commands like
> `echo`, `grep`, `test`, `printf` etc. When switching
> to non-POSIX compliant systems, one will have
> to re-implement these commands specifically for the
> system. There are also POSIX-to-Windows path conversion
> issues. To fix these issues, it was decided to convert
> these scripts into portable C code.
> "Large overhead in calling the command"
> The commands implemented in shell scripts are not builtins, so
> they call `fork()` and `exec()` multiple times, hence creating
> additional shells.


> This adds to the overhead in using the
> commands in terms of time and memory.
> "No access to low-level API"
> The shell commands don’t have access to low level commands
> like `git hash-object`, `git cat-file` etc.

I am not sure it's correct to say that they don't have access to such
low level commands. I think it would be better to say that they don't
have direct access to the internal functions behind such low level

> As these commands
> are internally required for submodule commands to work, the shell
> script needs to spawn a separate shell to execute these commands.

s/separate shell/separate process/

> =================================================
> ##How have I prepared?##
> I have gone through all the previous works and read through their
> code to make myself accustomed to the intricacies of the code.
> I have also structured my workflow based on the observation of
> the previous discussions on those patches, and taken into
> consideration the issues faced previously.
> =================================================
> ##Previous Work##
> A large part of the `git submodule--helper.c` has already been

I would say that it's '' that has been converted. 'git
submodule--helper.c' is the result of that conversion.

> converted by Stefan Beller, Prathamesh Chavan in his GSoC project
> in 2017, and Shourya Shukla in his GSoC project in 2020. This is
> the list of already ported commands.
> set-branch
> set-url
> summary
> status
> init
> deinit
> update
> foreach
> sync
> absorbgitdirs
> =================================================
> ##Work to be done##
> The only command that is left to be ported is `git submodule add`.
> The previous work on this by Shourya Shukla in GSoC 2020, did
> not reach a successful merge due to some issues in design and
> the patch was dropped as it had been stale for long.
> See:
> The first and foremost aim of the project will be to finish
> porting the `add` command. Thereafter, the end goal would be to
> completely replace the shell script ( with
> an efficient c code.


> Before porting the `git submodule add` command the initial work
> would be dedicated to the implementation of small helper functions
> in the form of small patches, which would be directly used by the
> `add` command. This workflow is based on the suggestion by
> Junio C Hamano on the thread:
> This workflow would help in the following ways:
> - It would help in sending patches in a small digestible format.
> - It would help the reviewers easily review those small units
>    of patches in a single sitting.
> - It would help keep small logical units of code in different clean commits.
> An additional test tweak would also be required in
> ``,
> to prepend the keyword ‘fatal’ since the command dies out in case
> of absence of commits as pointed out by Kaartic Sivaraam on the thread:


> The following helper functions would be required to be implemented -
> - A function to guess the path name from the repository string.
>    Example prototype: static char *guess_dir_name(const char *repo)
>    Returns the path name.
> - A function to assure repo is absolute or relative to parent.
>    Example prototype: static char *get_real_repo(const char *repo)
>    Returns the correct repo name.
> - A function for normalizing path, that is, removing multiple
>    //; leading ./; /./; /../; trailing / .
>    Example prototype: static char *normalize_path(const char *path)
> - A function to check if the path exists and is already a git
>    repo, else clone it.
>    Example prototype: static int add_submodule(struct add_data *info)
>    `add_data` is a struct which stores the config of the submodule.
> - A function to set the submodule config properly.
>    Example prototype: static void config_added_submodule(struct add_data
> *info)

How would you test these helper functions? Would they be used by any
code in 'builtin/submodule--helper.c'?

> - After implementation of all these helper methods, the main
>    `module_add()` function would be implemented using the helper
>    functions listed above as well as those helper functions which
>    are predefined.
> =================================================
> ##Project Timeline##
> "Present Day - May 17"
> I’ll utilize this time in exploring the codebase more properly and
> solving more issues, which would help me properly familiarize
> myself with the codebase. I’ll also try to structure a more
> solidified, detailed workflow and come up with a draft patch

A draft patch of what? I thought that you wanted to create many
patches, each one implementing a small step, maybe a single function
that you described above.

> based on the previous work and discussions.
> "May 17 - June 7 (Community bonding period)"
> - Get familiar with the community.
> - Discuss proper workflow with mentors.
> - Make changes in the timeline, if necessary.
> - Discuss the structure of the series of patches.
> "June 7 - June 25 (Initial coding phase)"
> - Finish implementation of the helper functions.
> - Work on a proper structure of the implementation of the
>    `submodule add` command

I am not sure what this means.

> and implement additional helper
>    functions if required.
> - Update proper documentation of the helper functions.
> "June 25 - July 5 (Binding the code)"
> This time would be used to code the main `submodule add`
> command using all the helper functions implemented in the
> initial phase of coding.

So how would the helper functions be useful and tested before this step?

> This includes binding all the code
> together and then completing the command through incremental
> reviews. Also, the necessary documentation would be updated
> parallelly.
> "July 5 - July 12 (Initiate porting of command)"
> - Discuss how to go about porting the entire submodule script.
> - Initiate porting of the `` script.
> "July 12 - July 16 (Phase 1 evaluation)"
> "July 16 - July 26 (Semester exams)"
> I will be taking my semester examinations during this
> time. As such, I’ll try to be in touch with the mentors
> and take out as much time as possible (around 20 hours a week).
> "July 26 - August 10 (Porting the complete script)"
> This period would be utilized in the complete conversion of
> `` into c code and combine it with

s/c code/C code/

> `submodule--helper.c` to make a single `builtin/submodule.c`.
> As I’ll be completely free from academics during this period,
> I’ll try to compensate as much time as possible for the above
> period of July 16 - July 26.

  reply	other threads:[~2021-04-10 12:04 UTC|newest]

Thread overview: 10+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2021-04-01 11:40 Chinmoy Chakraborty
2021-04-02  5:32 ` Bagas Sanjaya
2021-04-02  6:31   ` Christian Couder
2021-04-02  7:05     ` Chinmoy Chakraborty
2021-04-03 14:14     ` Chinmoy Chakraborty
2021-04-05 14:44       ` Christian Couder
2021-04-06 19:53         ` Chinmoy Chakraborty
2021-04-08  9:11         ` Chinmoy Chakraborty
2021-04-10 12:03           ` Christian Couder [this message]
2021-04-02  7:04   ` Chinmoy Chakraborty

Reply instructions:

You may reply publicly to this message via plain-text email
using any one of the following methods:

* Save the following mbox file, import it into your mail client,
  and reply-to-all from there: mbox

  Avoid top-posting and favor interleaved quoting:

  List information:

* Reply using the --to, --cc, and --in-reply-to
  switches of git-send-email(1):

  git send-email \
    --in-reply-to='' \ \ \ \ \
    --subject='Re: [GSoC] Draft Proposal (Convert submodule to builtin)' \

* If your mail client supports setting the In-Reply-To header
  via mailto: links, try the mailto: link

Code repositories for project(s) associated with this inbox:

This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for read-only IMAP folder(s) and NNTP newsgroup(s).