[GSoC] Intro and Micro-project

Hello, I am Edwin, a first year undergrad in the department of computing, Imperial College London. This is my first message to this mailing list! I want to work on the micro-project "avoiding pipelines in test scripts". I'm really excited to work on such a massive project (ie, git, not the microproject) and be in contact with the people who put it all together. I'm curious about how software development works at this scale, since it feels quite different from the small projects I've used to work on. I use bash quite frequently for automating tasks, and I have experience in C, through a uni project of making an assembler.

I found and looked at a few places with pipes. I have a few thoughts and questions on making a change. Firstly (if this is relevant), how do I ensure that the file I write the std output to doesn't have a name clash with other such files made during tests?

I will follow up with an email of the proposed patch. Regards, Edwin.

On Tue, Mar 21, 2023 at 9:32 AM Edwin Fernando <edwinfernando734@gmail.com> wrote:

Every test script is run inside a different directory. You can check the $PWD.

I'm not sure of the usefulness of such patch.

Cheers.

Hi Edwin!

On Tue, Mar 21, 2023 at 02:38:34PM +0000, Edwin Fernando wrote:

That sounds great! I assume that you meant avoiding pipelines where the 'git' executable is in a non-terminal position of the pipe, e.g.:

    $ git blah | <something else>

Since if "git blah" exited with a non-zero code or crashed, etc., then we wouldn't see the failure since the pipeline would suppress it.

That has been a long-standing goal within the test suite, and I think that it's a great project to get you started. It'll ensure that you have all of the bits in the right place to get Git running on your machine and that you're able to run the tests.

:-).

Each of the tests runs in their own "trash directory", which you can keep around after the tests are finished running if you invoke them with the '-d' flag. More information about each of those flags is available in t/README, as is information about the test suite as a whole.

I would also recommend that you take a look through t/test-lib.sh and t/test-lib-functions.sh, as those give a good overview of both (a) some of the internals of our integration testing library, and (b) some of the convenience functions available to you as a test author.

Looking at some example tests in the tree may be useful, too. You might want to check out t/t5318-commit-graph.sh as an example of our modern test-writing style.

Let us know if you have any questions or want any pointers on getting started.

Thanks, Taylor

Taylor Blau <me@ttaylorr.com> writes:

Yes, but can somebody rewrite the micro-project idea page to clarify what the "pipe" thing is about a bit more, so that you do not have to repeat the above explanation the next time ;-)?

It is not "we do not want pipe", but it is about "we do not want to ignore exit status of git". So just like

	$ git <subcommand> | <some other command>

sequence is bad, we want to avoid 

	$ <some command> $(git <subcommand>)

that feeds output of Git as an argument to some other command.  And

	<some command> <<EOF
	... some text ...
	$(git <subcommand)
	EOF

is bad too.  But

	var=$(git <subcommand>)

is OK ;-).

On Tue, Mar 21, 2023 at 10:41:33AM -0700, Junio C Hamano wrote:

Good suggestion. I did so in the following pull request:

  https://github.com/git/git.github.io/pull/633

Thanks, Taylor