Re: [PATCH v3 4/4] maintenance: use Windows scheduled tasks

From: Eric Sunshine <sunshine@sunshineco.com>
To: Derrick Stolee <stolee@gmail.com>
Cc: Derrick Stolee via GitGitGadget <gitgitgadget@gmail.com>,
	Git List <git@vger.kernel.org>,
	Derrick Stolee <derrickstolee@github.com>,
	Derrick Stolee <dstolee@microsoft.com>
Subject: Re: [PATCH v3 4/4] maintenance: use Windows scheduled tasks
Date: Fri, 13 Nov 2020 16:40:40 -0500	[thread overview]
Message-ID: <CAPig+cRv_Swd4WcZDnyDxg4E1HdwdkfybjWiZsYwPFSxkhUz0w@mail.gmail.com> (raw)
In-Reply-To: <73c902db-40f5-2693-c350-9c04ba2751cf@gmail.com>

On Fri, Nov 13, 2020 at 4:32 PM Derrick Stolee <stolee@gmail.com> wrote:
> On 11/13/2020 3:44 PM, Eric Sunshine wrote:
> > On Fri, Nov 13, 2020 at 9:00 AM Derrick Stolee via GitGitGadget
> > <gitgitgadget@gmail.com> wrote:
> >> +test_expect_success MINGW 'start and stop Windows maintenance' '
> >> +       write_script print-args <<-\EOF &&
> >> +       echo $* >>args
> >> +       EOF
> >
> > Using `>>` here makes it harder to reason about the test than using
> > `>` would, especially since `>>` seems to be unnecessary in this case.
>
> Since we execute the GIT_TEST_CRONTAB executable multiple times, we
> need to use >> to log all three instances (and their order). Using ">args"
> would only capture the final call for the weekly schedule.
>
> On macOS, there are as many as six calls (three bootouts, three bootstraps).

Makes sense. Thanks.

> >> +       GIT_TEST_CRONTAB="/bin/sh print-args" git maintenance start &&
> >
> > Is it a requirement on Windows to mention /bin/sh here? Specifically,
> > I'm wondering why a simple ./print-args doesn't work. (It's especially
> > unclear since write_script() is used heavily in the test suite and it
> > seems to work well enough on Windows without specifying /bin/sh.)
>
> I landed on this after trying several attempts to get this to work,
> including "$(pwd)/print-args" and I'm not sure why it doesn't work
> in the Windows case. It is something to do with how I am executing
> the subcommand from within Git. I'm pretty sure this idea of "mocking"
> an executable through Git is relatively new, or at least rare

Just for clarification... You mentioned in response to my [3/4] review
that your accidentally-working write_script() only worked as expected
on Mac but not on Windows. When you arrived at this solution of
GIT_TEST_CRONTAB="/bin/sh ..." here, was that before or after you
fixed write_script() to take the script body from stdin?