Re: [PATCH v6 6/6] config: "git config --get-urlmatch" parses section.<url>.key

From: Kyle J. McKay <mackyle@gmail.com>
Date: Jul 31, 2013, 23:03 UTC
Message-ID: <62E17EDB-B24D-4F37-95F8-E2E26118E5E9@gmail.com>
In-Reply-To: <20130731224511.GA25882@sigill.intra.peff.net>

On Jul 31, 2013, at 15:45, Jeff King wrote:

Show 27 quoted lines

> On Wed, Jul 31, 2013 at 12:26:08PM -0700, Junio C Hamano wrote:
>
>> Using the same urlmatch_config_entry() infrastructure, add a new
>> mode "--get-urlmatch" to the "git config" command, to learn values
>> for the "virtual" two-level variables customized for the specific
>> URL.
>>
>>    git config [--<type>] --get-urlmatch <section>[.<key>] <url>
>
> Do we want something like this on top, to convert the third form of
> test-url-normalize into git-config calls?
>
> It would be nicer squashed in, but we the tests are added earlier in  
> the
> series than "--get-urlmatch", so we would have to rip the tests out of
> the earlier patches and have a "[PATCH 7/6] add tests for url
> normalizing".
>
> Two things to note about my test conversion:
>
>  1. Git-config expects pre-canonicalized variable names (so  
> http.noepsv
>     instead of "http.noEPSV"). I think the "git config --get" code  
> path
>     does this for the caller, so we should probably do the same for
>     "--get-urlmatch". And it is even easier here, because we know that
>     "http.noEPSV" does not contain a case-sensitive middle part. :)

The test was testing that too, which I think is a good thing. Your replacement does not test that. With a fix for --get-urlmatch as you mention above, the tests can check that again.

Show 9 quoted lines

>  2. I turned the many 'test "$(git foo)" = "bar"' invocations into a
>     wrapper function that uses test_cmp. This helped immensely with
>     debugging (1).
>
>     The wrapper is a little ugly. I do wonder if we actually need all
>     of these tests (i.e., it is not clear to me what different things
>     each is testing, and if it is not simply trying to exercise the
>     different variable names, which now all follow the same code path,
>     because git-config does not care about the particular names).

Each one tests a different item from the "$tc-n" config file to make sure that everything that's in each config file actually behaves as expected.

If we do this (and I don't really have any objection except for the point noted above), then the tests really need to move out from t5200 as they're not tied to the http operations anymore. Also the Makefile rule for test-url-normalize.c needs to be simplified since it won't need the extra options to make it link since it's no longer including http.c.

The README has this:

Show 12 quoted lines

> First digit tells the family:
>
>         0 - the absolute basics and global stuff
>         1 - the basic commands concerning database
>         2 - the basic commands concerning the working tree
>         3 - the other basic commands (e.g. ls-files)
>         4 - the diff commands
>         5 - the pull and exporting commands
>         6 - the revision tree commands (even e.g. merge-base)
>         7 - the porcelainish commands concerning the working tree
>         8 - the porcelainish commands concerning forensics
>         9 - the git tools

But the best choice does not immediately jump out at me. However, looking at the other tests that are there, I think perhaps 1307-config- url might be a reasonable choice.