Thread (5 messages) 5 messages, 4 authors, 2017-10-30

Re: [PATCH] t0000: check whether the shell supports the "local" keyword

From: Michael Haggerty <hidden>
Date: 2017-10-26 08:48:09

On 10/26/2017 10:40 AM, Jacob Keller wrote:
On Thu, Oct 26, 2017 at 1:28 AM, Eric Sunshine [off-list ref] wrote:
quoted
On Thu, Oct 26, 2017 at 4:18 AM, Michael Haggerty [off-list ref] wrote:
quoted
Add a test balloon to see if we get complaints from anybody who is
using a shell that doesn't support the "local" keyword. If so, this
test can be reverted. If not, we might want to consider using "local"
in shell code throughout the git code base.
I would guess that the number of people who actually run the Git test
suite is microscopic compared to the number of people who use Git
itself. It is not clear, therefore, that lack of reports of failure of
the new test would imply that "local" can safely be used throughout
the Git code base. At best, it might indicate that "local" can be used
in the tests.

Or, am I missing something?
I don't think you're missing anything. I think the idea here is: "do
any users who actively run the test suite care if we start using
local". I don't think the goal is to allow use of local in non-test
suite code. At least, that's not how I interpreted it.

Thus it's fine to be only as part of a test and see if anyone
complains, since the only people affected would be those which
actually run the test suite...

Changing our requirement for regular shell scripts we ship seems a lot
trickier to gauge.
Actually, I would hope that if this experiment is successful that we can
use "local" in production code, too.

The proper question isn't "what fraction of Git users run the test
suite?", because I agree with Eric that that is microscopic. The correct
question is "on what fraction of platforms where Git will be run has the
test suite been run by *somebody*?", and I think (I hope!) that that
fraction is quite high.

Really...if you are compiling Git on a platform that is so deviant or
archaic that it doesn't have a reasonable Shell, and you don't even
bother running the test suite, you kindof deserve your fate, don't you?

Michael
Keyboard shortcuts
hback out one level
jnext message in thread
kprevious message in thread
ldrill in
Escclose help / fold thread tree
?toggle this help