[1/3] t/lib-httpd: avoid using BSD's sed

Commit Message

Linus Arver via GitGitGadget Feb. 26, 2020, 8:09 p.m. UTC

From: Johannes Schindelin <[e-mail]>

Among other differences relative to GNU sed, BSD sed always ends its
output with a trailing newline, even if the input did not have such a
trailing newline.

Surprisingly, this makes three httpd-based tests fail on macOS: t5616,
t5702 and t5703. ("Surprisingly" because those tests have been around
for some time, but apparently nobody runs them on macOS with a working
Apache2 setup.)

The reason is that we use `sed` in those tests to filter the response of
the web server. Apart from the fact that we use GNU constructs (such as
using a space after the `c` command instead of a backslash and a
newline), we have another problem: BSD sed LF-only newlines while
webservers are supposed to use CR/LF ones.

Even worse, t5616 uses `sed` to replace a binary part of the response
with a new binary part (kind of hoping that the replaced binary part
does not contain a 0x0a byte which would be interpreted as a newline).

To that end, it calls on Perl to read the binary pack file and
hex-encode it, then calls on `sed` to prefix every hex digit pair with a
`\x` in order to construct the text that the `c` statement of the `sed`
invocation is supposed to insert. So we call Perl and sed to construct a
sed statement. The final nail in the coffin is that BSD sed does not
even interpret those `\x<hex>` constructs.

Let's just replace all of that by Perl snippets. With Perl, at least, we
do not have to deal with GNU vs BSD semantics, we do not have to worry
about unwanted trailing newlines, and we do not have to spawn commands
to construct arguments for other commands to be spawned (i.e. we can
avoid a whole lot of shell scripting complexity).

The upshot is that this fixes t5616, t5702 and t5703 on macOS with
Apache2.

Signed-off-by: Johannes Schindelin <[e-mail]>
---
 t/lib-httpd.sh                     |  2 +-
 t/lib-httpd/apache.conf            |  6 ++---
 t/lib-httpd/apply-one-time-perl.sh | 27 +++++++++++++++++++++
 t/lib-httpd/apply-one-time-sed.sh  | 24 -------------------
 t/t5537-fetch-shallow.sh           | 10 ++++----
 t/t5616-partial-clone.sh           | 38 +++++++++++++++++-------------
 t/t5702-protocol-v2.sh             | 12 +++++-----
 t/t5703-upload-pack-ref-in-want.sh |  6 ++---
 8 files changed, 66 insertions(+), 59 deletions(-)
 create mode 100644 t/lib-httpd/apply-one-time-perl.sh
 delete mode 100644 t/lib-httpd/apply-one-time-sed.sh

Comments

Ed Maste Feb. 26, 2020, 8:20 p.m. UTC | #1

On Wed, 26 Feb 2020 at 15:09, Johannes Schindelin via GitGitGadget
<[e-mail]> wrote:
>
> From: Johannes Schindelin <[e-mail]>
>
> Among other differences relative to GNU sed, BSD sed always ends its
> output with a trailing newline, even if the input did not have such a
> trailing newline.
>
> Surprisingly, this makes three httpd-based tests fail on macOS: t5616,
> t5702 and t5703. ("Surprisingly" because those tests have been around
> for some time, but apparently nobody runs them on macOS with a working
> Apache2 setup.)

Hmm, this is interesting - all tests (that are executed) are passing
on FreeBSD, in CI.

I tried on FreeBSD and do not see a trailing newline added; I'm not
sure how sed behaves on other BSDs. However, you probably want to
refer to macOS sed rather than BSD sed in the commit.

Junio C Hamano Feb. 26, 2020, 8:26 p.m. UTC | #2

"Johannes Schindelin via GitGitGadget" <[e-mail]>
writes:

> diff --git a/t/t5537-fetch-shallow.sh b/t/t5537-fetch-shallow.sh
> index 9e16512fe31..4f10057e9f1 100755
> --- a/t/t5537-fetch-shallow.sh
> +++ b/t/t5537-fetch-shallow.sh
> @@ -237,7 +237,7 @@ test_expect_success 'shallow fetches check connectivity before writing shallow f
>  	git -C "$REPO" config protocol.version 2 &&
>  	git -C client config protocol.version 2 &&
>  
> -	git -C client fetch --depth=2 "$HTTPD_URL/one_time_sed/repo" master:a_branch &&
> +	git -C client fetch --depth=2 "$HTTPD_URL/one_time_perl/repo" master:a_branch &&
>  
>  	# Craft a situation in which the server sends back an unshallow request
>  	# with an empty packfile. This is done by refetching with a shorter
> @@ -246,13 +246,13 @@ test_expect_success 'shallow fetches check connectivity before writing shallow f
>  	printf "$(test_oid sed)" \

Hmm, shouldn't the test-oid token "sed" whose value is set up in the
setup section of this test script also be renamed to "perl"?  Or, if
we are actively taking advantage of the fact that the syntax of the
replacement operator is the same between the languages, perhaps "sed"
is better renamed to something more language agnostic and reflects
the purpose/reason why we extend the packet header by two bytes with
the one-time munging process?

>  	       "$(git -C "$REPO" rev-parse HEAD)" \
>  	       "$(git -C "$REPO" rev-parse HEAD^)" \
> -	       >"$HTTPD_ROOT_PATH/one-time-sed" &&
> +	       >"$HTTPD_ROOT_PATH/one-time-perl" &&

Other than that, this step looked quite sensible.  Thanks.

Junio C Hamano Feb. 26, 2020, 10:22 p.m. UTC | #3

Junio C Hamano <[e-mail]> writes:

> "Johannes Schindelin via GitGitGadget" <[e-mail]>
> writes:
>
>> diff --git a/t/t5537-fetch-shallow.sh b/t/t5537-fetch-shallow.sh
>> index 9e16512fe31..4f10057e9f1 100755
>> --- a/t/t5537-fetch-shallow.sh
>> +++ b/t/t5537-fetch-shallow.sh
>> @@ -237,7 +237,7 @@ test_expect_success 'shallow fetches check connectivity before writing shallow f
>>  	git -C "$REPO" config protocol.version 2 &&
>>  	git -C client config protocol.version 2 &&
>>  
>> -	git -C client fetch --depth=2 "$HTTPD_URL/one_time_sed/repo" master:a_branch &&
>> +	git -C client fetch --depth=2 "$HTTPD_URL/one_time_perl/repo" master:a_branch &&
>>  
>>  	# Craft a situation in which the server sends back an unshallow request
>>  	# with an empty packfile. This is done by refetching with a shorter
>> @@ -246,13 +246,13 @@ test_expect_success 'shallow fetches check connectivity before writing shallow f
>>  	printf "$(test_oid sed)" \
>
> Hmm, shouldn't the test-oid token "sed" whose value is set up in the
> setup section of this test script also be renamed to "perl"?  Or, if
> we are actively taking advantage of the fact that the syntax of the
> replacement operator is the same between the languages, perhaps "sed"
> is better renamed to something more language agnostic and reflects
> the purpose/reason why we extend the packet header by two bytes with
> the one-time munging process?
>
>>  	       "$(git -C "$REPO" rev-parse HEAD)" \
>>  	       "$(git -C "$REPO" rev-parse HEAD^)" \
>> -	       >"$HTTPD_ROOT_PATH/one-time-sed" &&
>> +	       >"$HTTPD_ROOT_PATH/one-time-perl" &&
>
> Other than that, this step looked quite sensible.  Thanks.

Hmm, is it because you wanted to backport this down to 'maint'
(otherwise, your tests will start failing in a month) that you left
the "test_oid sed" thing untouched?  If so, that makes sort-of
sense.

I expect that the series will be rerolled, if only for s/BSD/macOS/
mentioned elsewhere in the thread, but in the meantime, I'll rebase
them on 'maint' "as a practice" while queuing.

Johannes Schindelin Feb. 27, 2020, 3:40 p.m. UTC | #4

Hi Ed,

On Wed, 26 Feb 2020, Ed Maste wrote:

> On Wed, 26 Feb 2020 at 15:09, Johannes Schindelin via GitGitGadget
> <[e-mail]> wrote:
> >
> > From: Johannes Schindelin <[e-mail]>
> >
> > Among other differences relative to GNU sed, BSD sed always ends its
> > output with a trailing newline, even if the input did not have such a
> > trailing newline.
> >
> > Surprisingly, this makes three httpd-based tests fail on macOS: t5616,
> > t5702 and t5703. ("Surprisingly" because those tests have been around
> > for some time, but apparently nobody runs them on macOS with a working
> > Apache2 setup.)
>
> Hmm, this is interesting - all tests (that are executed) are passing
> on FreeBSD, in CI.
>
> I tried on FreeBSD and do not see a trailing newline added; I'm not
> sure how sed behaves on other BSDs. However, you probably want to
> refer to macOS sed rather than BSD sed in the commit.

My bad. I looked at StackOverflow and there the claim was that all BSD
seds behave that way.

Of course, SO always lags behind by a couple years (although even such old
threads are often very useful), so it is possible that _old_ BSD sed
behaved that way.

In any case, I adjusted the commit message.

Related, I saw that Cirrus CI offers FreeBSD builds, maybe you'd be
interested in supporting that out of the box in
https://github.com/git/git?

Ciao,
Dscho

Johannes Schindelin Feb. 27, 2020, 3:42 p.m. UTC | #5

Hi Junio,

On Wed, 26 Feb 2020, Junio C Hamano wrote:

> Junio C Hamano <[e-mail]> writes:
>
> > "Johannes Schindelin via GitGitGadget" <[e-mail]>
> > writes:
> >
> >> diff --git a/t/t5537-fetch-shallow.sh b/t/t5537-fetch-shallow.sh
> >> index 9e16512fe31..4f10057e9f1 100755
> >> --- a/t/t5537-fetch-shallow.sh
> >> +++ b/t/t5537-fetch-shallow.sh
> >> @@ -237,7 +237,7 @@ test_expect_success 'shallow fetches check connectivity before writing shallow f
> >>  	git -C "$REPO" config protocol.version 2 &&
> >>  	git -C client config protocol.version 2 &&
> >>
> >> -	git -C client fetch --depth=2 "$HTTPD_URL/one_time_sed/repo" master:a_branch &&
> >> +	git -C client fetch --depth=2 "$HTTPD_URL/one_time_perl/repo" master:a_branch &&
> >>
> >>  	# Craft a situation in which the server sends back an unshallow request
> >>  	# with an empty packfile. This is done by refetching with a shorter
> >> @@ -246,13 +246,13 @@ test_expect_success 'shallow fetches check connectivity before writing shallow f
> >>  	printf "$(test_oid sed)" \
> >
> > Hmm, shouldn't the test-oid token "sed" whose value is set up in the
> > setup section of this test script also be renamed to "perl"?

Ooops...

> > Or, if we are actively taking advantage of the fact that the syntax of
> > the replacement operator is the same between the languages, perhaps
> > "sed" is better renamed to something more language agnostic and
> > reflects the purpose/reason why we extend the packet header by two
> > bytes with the one-time munging process?
> >
> >>  	       "$(git -C "$REPO" rev-parse HEAD)" \
> >>  	       "$(git -C "$REPO" rev-parse HEAD^)" \
> >> -	       >"$HTTPD_ROOT_PATH/one-time-sed" &&
> >> +	       >"$HTTPD_ROOT_PATH/one-time-perl" &&
> >
> > Other than that, this step looked quite sensible.  Thanks.
>
> Hmm, is it because you wanted to backport this down to 'maint'
> (otherwise, your tests will start failing in a month) that you left
> the "test_oid sed" thing untouched?  If so, that makes sort-of
> sense.

That's a good point. I target `maint` in v2, and offered an add-on patch
meant to be applied on top of the merge into `master` (or `next`, or
`pu`).

> I expect that the series will be rerolled, if only for s/BSD/macOS/
> mentioned elsewhere in the thread, but in the meantime, I'll rebase
> them on 'maint' "as a practice" while queuing.

Thanks ;-)

Ciao,
Dscho

Ed Maste Feb. 27, 2020, 5:39 p.m. UTC | #6

On Thu, 27 Feb 2020 at 10:40, Johannes Schindelin
<[e-mail]> wrote:
>
> My bad. I looked at StackOverflow and there the claim was that all BSD
> seds behave that way.
>
> Of course, SO always lags behind by a couple years (although even such old
> threads are often very useful), so it is possible that _old_ BSD sed
> behaved that way.

Yeah, I wondered about the different behaviour, and asked on Twitter
about it. You're right, it is historical BSD behaviour and NetBSD at
least still appends the newline. FreeBSD changed this in 2014 -
http://bugs.freebsd.org/160745.

> Related, I saw that Cirrus CI offers FreeBSD builds, maybe you'd be
> interested in supporting that out of the box in
> https://github.com/git/git?

Indeed - there is a .cirrus.yml in git now which builds and runs tests
(on FreeBSD 12.1). I'll look into working with the GitHub organization
owners for git and gitgitgadget to see about allowing Cirrus to access
the repositories.

Johannes Schindelin Feb. 27, 2020, 7:46 p.m. UTC | #7

Hi Ed,

On Thu, 27 Feb 2020, Ed Maste wrote:

> On Thu, 27 Feb 2020 at 10:40, Johannes Schindelin
> <[e-mail]> wrote:
> >
> > My bad. I looked at StackOverflow and there the claim was that all BSD
> > seds behave that way.
> >
> > Of course, SO always lags behind by a couple years (although even such old
> > threads are often very useful), so it is possible that _old_ BSD sed
> > behaved that way.
>
> Yeah, I wondered about the different behaviour, and asked on Twitter
> about it. You're right, it is historical BSD behaviour and NetBSD at
> least still appends the newline. FreeBSD changed this in 2014 -
> http://bugs.freebsd.org/160745.

Thank you for digging into this!

> > Related, I saw that Cirrus CI offers FreeBSD builds, maybe you'd be
> > interested in supporting that out of the box in
> > https://github.com/git/git?
>
> Indeed - there is a .cirrus.yml in git now which builds and runs tests
> (on FreeBSD 12.1). I'll look into working with the GitHub organization
> owners for git and gitgitgadget to see about allowing Cirrus to access
> the repositories.

Oy, I had forgotten that you worked on this. I enabled this in the git and
in the gitgitgadget orgs. The next pushes/PRs should benefit from this.

Thanks,
Dscho

Johannes Schindelin Feb. 28, 2020, 11:57 p.m. UTC | #8

Hi Ed,

On Thu, 27 Feb 2020, Johannes Schindelin wrote:

> On Thu, 27 Feb 2020, Ed Maste wrote:
>
> > On Thu, 27 Feb 2020 at 10:40, Johannes Schindelin
> > <[e-mail]> wrote:
> >
> > > Related, I saw that Cirrus CI offers FreeBSD builds, maybe you'd be
> > > interested in supporting that out of the box in
> > > https://github.com/git/git?
> >
> > Indeed - there is a .cirrus.yml in git now which builds and runs tests
> > (on FreeBSD 12.1). I'll look into working with the GitHub organization
> > owners for git and gitgitgadget to see about allowing Cirrus to access
> > the repositories.
>
> Oy, I had forgotten that you worked on this. I enabled this in the git
> and in the gitgitgadget orgs. The next pushes/PRs should benefit from
> this.

This works now. When you click on the green checkmarks (and the one red X)
at https://github.com/git/git/branches/active, you will see the
`freebsd_12` build.

Ciao,
Dscho

Patch

diff --git a/t/lib-httpd.sh b/t/lib-httpd.sh
index 656997b4d66..1449ee95e9e 100644
--- a/t/lib-httpd.sh
+++ b/t/lib-httpd.sh
@@ -132,7 +132,7 @@  prepare_httpd() {
 	install_script broken-smart-http.sh
 	install_script error-smart-http.sh
 	install_script error.sh
-	install_script apply-one-time-sed.sh
+	install_script apply-one-time-perl.sh
 
 	ln -s "$LIB_HTTPD_MODULE_PATH" "$HTTPD_ROOT_PATH/modules"
 
diff --git a/t/lib-httpd/apache.conf b/t/lib-httpd/apache.conf
index 5c1c86c193a..994e5290d63 100644
--- a/t/lib-httpd/apache.conf
+++ b/t/lib-httpd/apache.conf
@@ -113,7 +113,7 @@  Alias /auth/dumb/ www/auth/dumb/
 	SetEnv GIT_EXEC_PATH ${GIT_EXEC_PATH}
 	SetEnv GIT_HTTP_EXPORT_ALL
 </LocationMatch>
-<LocationMatch /one_time_sed/>
+<LocationMatch /one_time_perl/>
 	SetEnv GIT_EXEC_PATH ${GIT_EXEC_PATH}
 	SetEnv GIT_HTTP_EXPORT_ALL
 </LocationMatch>
@@ -122,7 +122,7 @@  ScriptAliasMatch /smart_*[^/]*/(.*) ${GIT_EXEC_PATH}/git-http-backend/$1
 ScriptAlias /broken_smart/ broken-smart-http.sh/
 ScriptAlias /error_smart/ error-smart-http.sh/
 ScriptAlias /error/ error.sh/
-ScriptAliasMatch /one_time_sed/(.*) apply-one-time-sed.sh/$1
+ScriptAliasMatch /one_time_perl/(.*) apply-one-time-perl.sh/$1
 <Directory ${GIT_EXEC_PATH}>
 	Options FollowSymlinks
 </Directory>
@@ -135,7 +135,7 @@  ScriptAliasMatch /one_time_sed/(.*) apply-one-time-sed.sh/$1
 <Files error.sh>
   Options ExecCGI
 </Files>
-<Files apply-one-time-sed.sh>
+<Files apply-one-time-perl.sh>
 	Options ExecCGI
 </Files>
 <Files ${GIT_EXEC_PATH}/git-http-backend>
diff --git a/t/lib-httpd/apply-one-time-perl.sh b/t/lib-httpd/apply-one-time-perl.sh
new file mode 100644
index 00000000000..09a0abdff7c
--- /dev/null
+++ b/t/lib-httpd/apply-one-time-perl.sh
@@ -0,0 +1,27 @@ 
+#!/bin/sh
+
+# If "one-time-perl" exists in $HTTPD_ROOT_PATH, run perl on the HTTP response,
+# using the contents of "one-time-perl" as the perl command to be run. If the
+# response was modified as a result, delete "one-time-perl" so that subsequent
+# HTTP responses are no longer modified.
+#
+# This can be used to simulate the effects of the repository changing in
+# between HTTP request-response pairs.
+if test -f one-time-perl
+then
+	LC_ALL=C
+	export LC_ALL
+
+	"$GIT_EXEC_PATH/git-http-backend" >out
+	perl -pe "$(cat one-time-perl)" out >out_modified
+
+	if cmp -s out out_modified
+	then
+		cat out
+	else
+		cat out_modified
+		rm one-time-perl
+	fi
+else
+	"$GIT_EXEC_PATH/git-http-backend"
+fi
diff --git a/t/lib-httpd/apply-one-time-sed.sh b/t/lib-httpd/apply-one-time-sed.sh
deleted file mode 100644
index bf7689d0202..00000000000
--- a/t/lib-httpd/apply-one-time-sed.sh
+++ /dev/null
@@ -1,24 +0,0 @@ 
-#!/bin/sh
-
-# If "one-time-sed" exists in $HTTPD_ROOT_PATH, run sed on the HTTP response,
-# using the contents of "one-time-sed" as the sed command to be run. If the
-# response was modified as a result, delete "one-time-sed" so that subsequent
-# HTTP responses are no longer modified.
-#
-# This can be used to simulate the effects of the repository changing in
-# between HTTP request-response pairs.
-if test -f one-time-sed
-then
-	"$GIT_EXEC_PATH/git-http-backend" >out
-	sed "$(cat one-time-sed)" out >out_modified
-
-	if cmp -s out out_modified
-	then
-		cat out
-	else
-		cat out_modified
-		rm one-time-sed
-	fi
-else
-	"$GIT_EXEC_PATH/git-http-backend"
-fi
diff --git a/t/t5537-fetch-shallow.sh b/t/t5537-fetch-shallow.sh
index 9e16512fe31..4f10057e9f1 100755
--- a/t/t5537-fetch-shallow.sh
+++ b/t/t5537-fetch-shallow.sh
@@ -237,7 +237,7 @@  test_expect_success 'shallow fetches check connectivity before writing shallow f
 	git -C "$REPO" config protocol.version 2 &&
 	git -C client config protocol.version 2 &&
 
-	git -C client fetch --depth=2 "$HTTPD_URL/one_time_sed/repo" master:a_branch &&
+	git -C client fetch --depth=2 "$HTTPD_URL/one_time_perl/repo" master:a_branch &&
 
 	# Craft a situation in which the server sends back an unshallow request
 	# with an empty packfile. This is done by refetching with a shorter
@@ -246,13 +246,13 @@  test_expect_success 'shallow fetches check connectivity before writing shallow f
 	printf "$(test_oid sed)" \
 	       "$(git -C "$REPO" rev-parse HEAD)" \
 	       "$(git -C "$REPO" rev-parse HEAD^)" \
-	       >"$HTTPD_ROOT_PATH/one-time-sed" &&
+	       >"$HTTPD_ROOT_PATH/one-time-perl" &&
 	test_must_fail env GIT_TEST_SIDEBAND_ALL=0 git -C client \
-		fetch --depth=1 "$HTTPD_URL/one_time_sed/repo" \
+		fetch --depth=1 "$HTTPD_URL/one_time_perl/repo" \
 		master:a_branch &&
 
-	# Ensure that the one-time-sed script was used.
-	! test -e "$HTTPD_ROOT_PATH/one-time-sed" &&
+	# Ensure that the one-time-perl script was used.
+	! test -e "$HTTPD_ROOT_PATH/one-time-perl" &&
 
 	# Ensure that the resulting repo is consistent, despite our failure to
 	# fetch.
diff --git a/t/t5616-partial-clone.sh b/t/t5616-partial-clone.sh
index 9a9178fd281..0eb5b1c47b1 100755
--- a/t/t5616-partial-clone.sh
+++ b/t/t5616-partial-clone.sh
@@ -398,14 +398,18 @@  intersperse () {
 	sed 's/\(..\)/'$1'\1/g'
 }
 
-# Create a one-time-sed command to replace the existing packfile with $1.
+# Create a one-time-perl command to replace the existing packfile with $1.
 replace_packfile () {
 	# The protocol requires that the packfile be sent in sideband 1, hence
 	# the extra \x01 byte at the beginning.
-	printf "1,/packfile/!c %04x\\\\x01%s0000" \
-		"$(($(wc -c <$1) + 5))" \
-		"$(hex_unpack <$1 | intersperse '\\x')" \
-		>"$HTTPD_ROOT_PATH/one-time-sed"
+	cp $1 "$HTTPD_ROOT_PATH/one-time-pack" &&
+	echo 'if (/packfile/) {
+		print;
+		my $length = -s "one-time-pack";
+		printf "%04x\x01", $length + 5;
+		print `cat one-time-pack` . "0000";
+		last
+	}' >"$HTTPD_ROOT_PATH/one-time-perl"
 }
 
 test_expect_success 'upon cloning, check that all refs point to objects' '
@@ -429,16 +433,16 @@  test_expect_success 'upon cloning, check that all refs point to objects' '
 	# \x01 byte at the beginning.
 	replace_packfile incomplete.pack &&
 
-	# Use protocol v2 because the sed command looks for the "packfile"
+	# Use protocol v2 because the perl command looks for the "packfile"
 	# section header.
 	test_config -C "$SERVER" protocol.version 2 &&
 	test_must_fail git -c protocol.version=2 clone \
-		--filter=blob:none $HTTPD_URL/one_time_sed/server repo 2>err &&
+		--filter=blob:none $HTTPD_URL/one_time_perl/server repo 2>err &&
 
 	test_i18ngrep "did not send all necessary objects" err &&
 
-	# Ensure that the one-time-sed script was used.
-	! test -e "$HTTPD_ROOT_PATH/one-time-sed"
+	# Ensure that the one-time-perl script was used.
+	! test -e "$HTTPD_ROOT_PATH/one-time-perl"
 '
 
 test_expect_success 'when partial cloning, tolerate server not sending target of tag' '
@@ -469,17 +473,17 @@  test_expect_success 'when partial cloning, tolerate server not sending target of
 	# \x01 byte at the beginning.
 	replace_packfile incomplete.pack &&
 
-	# Use protocol v2 because the sed command looks for the "packfile"
+	# Use protocol v2 because the perl command looks for the "packfile"
 	# section header.
 	test_config -C "$SERVER" protocol.version 2 &&
 
 	# Exercise to make sure it works.
 	git -c protocol.version=2 clone \
-		--filter=blob:none $HTTPD_URL/one_time_sed/server repo 2> err &&
+		--filter=blob:none $HTTPD_URL/one_time_perl/server repo 2> err &&
 	! grep "missing object referenced by" err &&
 
-	# Ensure that the one-time-sed script was used.
-	! test -e "$HTTPD_ROOT_PATH/one-time-sed"
+	# Ensure that the one-time-perl script was used.
+	! test -e "$HTTPD_ROOT_PATH/one-time-perl"
 '
 
 test_expect_success 'tolerate server sending REF_DELTA against missing promisor objects' '
@@ -502,7 +506,7 @@  test_expect_success 'tolerate server sending REF_DELTA against missing promisor
 
 	# Clone. The client has deltabase_have but not deltabase_missing.
 	git -c protocol.version=2 clone --no-checkout \
-		--filter=blob:none $HTTPD_URL/one_time_sed/server repo &&
+		--filter=blob:none $HTTPD_URL/one_time_perl/server repo &&
 	git -C repo hash-object -w -- "$SERVER/have.txt" &&
 
 	# Sanity check to ensure that the client does not have
@@ -543,7 +547,7 @@  test_expect_success 'tolerate server sending REF_DELTA against missing promisor
 
 	replace_packfile thin.pack &&
 
-	# Use protocol v2 because the sed command looks for the "packfile"
+	# Use protocol v2 because the perl command looks for the "packfile"
 	# section header.
 	test_config -C "$SERVER" protocol.version 2 &&
 
@@ -556,8 +560,8 @@  test_expect_success 'tolerate server sending REF_DELTA against missing promisor
 	grep "want $(cat deltabase_missing)" trace &&
 	! grep "want $(cat deltabase_have)" trace &&
 
-	# Ensure that the one-time-sed script was used.
-	! test -e "$HTTPD_ROOT_PATH/one-time-sed"
+	# Ensure that the one-time-perl script was used.
+	! test -e "$HTTPD_ROOT_PATH/one-time-perl"
 '
 
 # DO NOT add non-httpd-specific tests here, because the last part of this
diff --git a/t/t5702-protocol-v2.sh b/t/t5702-protocol-v2.sh
index 7fd7102c874..5039e66dc47 100755
--- a/t/t5702-protocol-v2.sh
+++ b/t/t5702-protocol-v2.sh
@@ -712,11 +712,11 @@  test_expect_success 'when server sends "ready", expect DELIM' '
 
 	# After "ready" in the acknowledgments section, pretend that a FLUSH
 	# (0000) was sent instead of a DELIM (0001).
-	printf "/ready/,$ s/0001/0000/" \
-		>"$HTTPD_ROOT_PATH/one-time-sed" &&
+	printf "\$ready = 1 if /ready/; \$ready && s/0001/0000/" \
+		>"$HTTPD_ROOT_PATH/one-time-perl" &&
 
 	test_must_fail git -C http_child -c protocol.version=2 \
-		fetch "$HTTPD_URL/one_time_sed/http_parent" 2> err &&
+		fetch "$HTTPD_URL/one_time_perl/http_parent" 2> err &&
 	test_i18ngrep "expected packfile to be sent after .ready." err
 '
 
@@ -737,12 +737,12 @@  test_expect_success 'when server does not send "ready", expect FLUSH' '
 
 	# After the acknowledgments section, pretend that a DELIM
 	# (0001) was sent instead of a FLUSH (0000).
-	printf "/acknowledgments/,$ s/0000/0001/" \
-		>"$HTTPD_ROOT_PATH/one-time-sed" &&
+	printf "\$ack = 1 if /acknowledgments/; \$ack && s/0000/0001/" \
+		>"$HTTPD_ROOT_PATH/one-time-perl" &&
 
 	test_must_fail env GIT_TRACE_PACKET="$(pwd)/log" git -C http_child \
 		-c protocol.version=2 \
-		fetch "$HTTPD_URL/one_time_sed/http_parent" 2> err &&
+		fetch "$HTTPD_URL/one_time_perl/http_parent" 2> err &&
 	grep "fetch< .*acknowledgments" log &&
 	! grep "fetch< .*ready" log &&
 	test_i18ngrep "expected no other sections to be sent after no .ready." err
diff --git a/t/t5703-upload-pack-ref-in-want.sh b/t/t5703-upload-pack-ref-in-want.sh
index 8aeeaac5091..7fba3063bf9 100755
--- a/t/t5703-upload-pack-ref-in-want.sh
+++ b/t/t5703-upload-pack-ref-in-want.sh
@@ -314,7 +314,7 @@  test_expect_success 'setup repos for change-while-negotiating test' '
 		test_commit m3 &&
 		git tag -d m2 m3
 	) &&
-	git -C "$LOCAL_PRISTINE" remote set-url origin "http://127.0.0.1:$LIB_HTTPD_PORT/one_time_sed/repo" &&
+	git -C "$LOCAL_PRISTINE" remote set-url origin "http://127.0.0.1:$LIB_HTTPD_PORT/one_time_perl/repo" &&
 	git -C "$LOCAL_PRISTINE" config protocol.version 2
 '
 
@@ -327,7 +327,7 @@  inconsistency () {
 	# RPCs during a single negotiation.
 	oid1=$(git -C "$REPO" rev-parse $1) &&
 	oid2=$(git -C "$REPO" rev-parse $2) &&
-	echo "s/$oid1/$oid2/" >"$HTTPD_ROOT_PATH/one-time-sed"
+	echo "s/$oid1/$oid2/" >"$HTTPD_ROOT_PATH/one-time-perl"
 }
 
 test_expect_success 'server is initially ahead - no ref in want' '
@@ -379,7 +379,7 @@  test_expect_success 'server loses a ref - ref in want' '
 	git -C "$REPO" config uploadpack.allowRefInWant true &&
 	rm -rf local &&
 	cp -r "$LOCAL_PRISTINE" local &&
-	echo "s/master/raster/" >"$HTTPD_ROOT_PATH/one-time-sed" &&
+	echo "s/master/raster/" >"$HTTPD_ROOT_PATH/one-time-perl" &&
 	test_must_fail git -C local fetch 2>err &&
 
 	test_i18ngrep "fatal: remote error: unknown ref refs/heads/raster" err