Re: Drastic jump in the time required for the test suite

[Jeff King <peff@peff.net>]

On Thu, Oct 20, 2016 at 12:54:32PM -0700, Stefan Beller wrote:

> Maybe we should stop introducing un-optimized tests.
> [...]
> * heavy use of the "git -C <dir>" pattern. When applying that
>   thouroughly we'd save spanning the subshells.

Yeah, I imagine with some style changes we could drop quite a few
subshells. The problem is that the conversion work is manual and
tedious. I'd look first for spots where we can eliminate thousands of
calls with a single change.

> That said I really like the idea of having a helper that would eliminate the cat
> for you, e.g. :
> 
> git_test_helper_equal_stdin_or_diff_and_die -C super_repo status
> --porcelain=v2 --branch --untracked-files=all <<-EOF
> 1 A. N... 000000 100644 100644 $_z40 $HMOD .gitmodules
> 1 AM S.M. 000000 160000 160000 $_z40 $HSUP sub1
> EOF

I think that helper still ends up using "cat" and "diff" under the hood,
unless you write those bits in pure shell. But at that point, I suspect
we could "cat" and "test_cmp" in pure shell, something like:

	cat () {
		# optimize common here-doc usage
		if test $# -eq 0
		then
			while read -r line
			do
				printf '%s' "$line"
			done
		fi
		command cat "$@"
	}

	test_cmp () {
		# optimize for common "they are the same" case
		# without any subshells or subprograms
		while true; do
			if ! read -r line1 <&3
			then
				if ! read -r line2 <&4
					# EOF on both; good
					return 0
				else
					# EOF only on file1; fail
					break
				fi
			fi
			if ! read -r line2 <&4
			then
				# EOF only on file2; fail
				break
			fi
			test "$line1" = "$line2" || break
		done 3<"$1" 4<"$2"

		# if we get here, the optimized version found some
		# difference. We can just "return 1", but let's run
		# the real $GIT_TEST_CMP to provide pretty output.
		# This should generally only happen on test failures,
		# so performance isn't a big deal.
		"$GIT_TEST_CMP" "$@"
	}

Those are both completely untested. But maybe they are worth playing
around with for somebody on Windows to see if they make a dent in the
test runtime.

-Peff