Re: [PATCH 1/2] t0302: check helper can handle empty credentials

From: Johannes Schindelin <johannes.schindelin@gmx.de>
Date: Oct 31, 2017, 17:16 UTC
Message-ID: <alpine.DEB.2.21.1.1710311814260.6482@virtualbox>
In-Reply-To: <20171030182718.oyzhjffd6jd4wslu@sigill.intra.peff.net>

Hi Peff,

On Mon, 30 Oct 2017, Jeff King wrote:

Show 20 quoted lines

> On Mon, Oct 30, 2017 at 06:20:12PM +0100, Johannes Schindelin wrote:
> 
> > Subject: Re: [PATCH 1/2] t0302: check helper can handle empty credentials
> 
> I guess we really care about t0303 here (which tests external helpers).
> This patch adds the test to lib-credential, so it hits the "cache" and
> "store" helpers, too. Which seems to pass, so I guess that's OK (I have
> to admit that as the author of those tools, I wasn't sure how they'd
> react).
> 
> > Make sure the helper does not crash when blank username and password is
> > provided. If the helper can save such credentials, it should be able to
> > read them back.
> 
> I worry that some third-party helpers might not be able to represent
> this case and would fail the test. This has been around for years no
> Windows, but probably hasn't ever been run with osxkeychain or
> libsecret. I'd be OK with taking this as-is, though, and waiting to see
> if anybody complains. At that point we'll know if the right solution is
> enhancing that helper, or providing a way to optionally skip this test.

Okay. If you change your mind, please let me know, I would try to set aside some time to adjust the patch in that event.

> (Though I have no idea if anybody actually runs t0303 against
> custom-built helpers in the first place. The process is pretty manual
> for now, though the Makefiles in contrib/credential could probably at
> least provide a "make test").

Right... I am not aware of any attempts to run those tests against Git Credential Manager for Windows, for example...

Ciao, Dscho