Re: [PATCH v3 3/3] connect: improve check for plink to reduce false positives

[Jeff King <peff@peff.net>]

On Sun, Apr 26, 2015 at 08:30:12PM +0000, brian m. carlson wrote:

> The git_connect function has code to handle plink and tortoiseplink
> specially, as they require different command line arguments from
> OpenSSH (-P instead of -p for ports; tortoiseplink additionally requires
> -batch).  However, the match was done by checking for "plink" anywhere
> in the string, which led to a GIT_SSH value containing "uplink" being
> treated as an invocation of putty's plink.
> 
> Improve the check by looking for "plink" or "tortoiseplink" (or those
> names suffixed with ".exe") only in the final component of the path.
> This has the downside that a program such as "plink-0.63" would no
> longer be recognized, but the increased robustness is likely worth it.
> Add tests to cover these cases to avoid regressions.

FYI, this ended up biting me today. We have some integration tests that
make sure we can clone over putty, and we wrap plink in a
"plink-wrapper.sh" script that tweaks a few extra options. That used to
match under the old scheme, but not the new. It would also match if we
looked for "plink" anywhere in the basename (but not in leading
directories).

I was able to work around it pretty easily by changing our test setup,
but I thought I would include it here as a data point. It's probably not
that representative of real-world users.

-Peff