GSoC 2016 Microproject

Hi everyone,

I'm Sidhant Sharma. I'm looking to participate in GSoC 2016 (and contribute to Git in general as well). I read up the pages relating to participation in GSoC and selected the microproject "Add configuration options for some commonly used command-line options". I have some queries regarding this though:

1. I believe the option can be added by a reassignment of the `verbose`  
variable in `parse_and_validate_options` function inside the  
`builtin/commit.c`. It can be a logical OR between the `verbose` (which was  
parsed by the argument parser) and the value obtained by using  
`git_config_get_bool("command.verbose", &opt_verbose)`. Am I thinking on  
the right lines?

2. Should I also add an option to override the configuration for verbose?

3. The microproject title suggests there can more such options, so are  
there more to be added?

Thanks in advance.

Regards

Sidhant Sharma [:tk]

Sidhant Sharma <tigerkid001@gmail.com> writes:

There's already work in progress on this project, see the mailing list the last few days.

Okay, in that case, I'd like to try is the one titled "Make upload-pack and receive-pack use the parse-options api". In that, I can take up the `builtin/receive-pack.c` file. Would that be okay?

Thanks again.

Regards, Sidhant Sharma [:tk]

Sidhant Sharma <tigerkid001@gmail.com> writes:

I didn't see anything going on for this one, but you may want to double-check with the ml's archives.

Hi,

I checked the archives and there doesn't seem to be any active work on this. I made the required changes and ran the test suite. Though all the tests pass, there still are two queries I have. First, I'm not quite sure what to put in the help message for the options (--quiet, --stateless-rpc, --advertise-refs and --reject-thin-pack-for-testing). Second, regarding the reject-thin-pack-for-testing option, currently when the option is entered, `fix_thin` is unset (https://github.com/git/git/blob/master/builtin/receive-pack.c#L1736). But using `OPT_BOOL(...)` for the same, the variable would instead be set when the option is given. I think one solution can be to invert `fix_thin` after calling `parse_options`. Am I going right so far? Suggestions and corrections welcome.

Regards, Sidhant Sharma [:tk]

Hi, Should I make a patch for this and submit it for discussion on the mailing list?

Regards, Sidhant Sharma [:tk]

Sidhant Sharma <tigerkid001@gmail.com> writes:

They are currently undocumented. We sometimes have explicitly undocumented options (PARSE_OPT_HIDDEN) when they are used only internally to avoid polluting the end-user's UI.

In this case, the command is anyway not meant for end-users so I think it would make sense to document them, but not necessarily within the the microproject.

Or use OPT_SET_INT(..., 0) on a variable initialized to 1.

On this list, it is indeed often more efficient to say "here's what I'm done. Any comments?" than "here's what I'm about to do".

So what may I put in the message parameter? I was thinking perhaps the option itself, without hyphens. Would that be correct?

Okay, will do that.

I'm really sorry, I'm not very familiar with mailing list etiquettes. I'll keep that in mind :)

Thanks and regards, Sidhant Sharma [:tk]

Sidhant Sharma <tigerkid001@gmail.com> writes:

If you use PARSE_OPT_HIDDEN, I think you don't need to specify a message. Otherwise, the documentation only has value if it contains more than just the option name, but that is the hard part if you're not familiar with the code. The best place to find documentation is in the history (git blame the file and see if the commit message introducing the option enlightens you). But that's why I said this didn't have to be part of the microproject: writting good doc requires a good understanding of the whole thing ...

No problem. It's OK to say what you do beforehand and to ask help. Just don't be surprised when you don't get much feedback on message not starting with [PATCH] ;-).

I used OPT_HIDDEN_BOOL for all except for reject-thin-pack-for-testing, where I used PARSE_OPT_HIDDEN. I ran the test locally and also on Travis, and the all tests passed. How do I proceed now?

Thanks and regards, Sidhant Sharma [:tk]

On 03/01, Sidhant Sharma wrote:

Now you can send a patch to the mailing list. See Documentation/SubmittingPatches for more details.

It usually is helpful to send the patches to yourself first as well, and try to apply them using git am, to avoid mistakes in the patch submission.

-- Thomas