[PATCH v2 1/5] test: avoid showing legacy-display parts

Subject: [PATCH v2 1/5] test: avoid showing legacy-display parts

Date: Fri, 31 May 2019 03:59:03 -0400

To: Notmuch Mail


From: Daniel Kahn Gillmor

Enigmail generates a "legacy-display" part when it sends encrypted
mail with a protected Subject: header.  This part is intended to
display the Subject for mail user agents that are capable of
decryption, but do not know how to deal with embedded protected

This part is the first child of a two-part multipart/mixed
cryptographic payload within a cryptographic envelope that includes
encryption (that is, it is not just a cleartext signed message).  It
uses Content-Type: text/rfc822-headers.

That is:

A └┬╴multipart/encrypted
B  ├─╴application/pgp-encrypted
C  └┬╴application/octet-stream
*   ╤ <decryption>
D   └┬╴multipart/mixed; protected-headers=v1 (cryptographic payload)
E    ├─╴text/rfc822-headers; protected-headers=v1 (legacy-display part)
F    └─╴… (actual message body)

In discussions with jrollins, i've come to the conclusion that a
legacy-display part should be stripped entirely from "notmuch show"
and "notmuch reply" now that these tools can understand and interpret
protected headers.

You can tell when a message part is a protected header part this way:

 * is the payload (D) multipart/mixed with exactly two children?
 * is its first child (E) Content-Type: text/rfc822-headers?
 * does the first child (E) have the property protected-headers=v1?
 * do all the headers in the body of the first child (E) match
   the protected headers in the payload part (D) itself?

If this is the case, and we already know how to deal with the
protected header, then there is no reason to try to render the
legacy-display part itself for the user.

Furthermore, when indexing, if we are indexing properly, we should
avoid indexing the text in E as part of the message body.

'notmuch reply' is an interesting case: the standard use of 'notmuch
reply' will end up omitting all mention of protected Subject:.

The right fix is for the replying MUA to be able to protect its
headers, and for it to set them appropriately based on headers found
in the original message.

If a replying MUA is unable to protect headers, but still wants the
user to be able to see the original header, a replying MUA that
notices that the original message's subject differs from the proposed
reply subject may choose to include the original's subject in the
quoted/attributed text. (this would be a stopgap measure; it's not
even clear that there is user demand for it)

This test suite change indicates what we want to happen for this case
(the tests are currently broken), and includes three additional TODO
suggestions of subtle cases for anyone who wants to flesh out the test
suite even further.  (i believe all these cases should be already
fixed by the rest of this series, but haven't had time to write the
tests for the unusual cases)

Signed-off-by: Daniel Kahn Gillmor <dkg@fifthhorseman.net>
 test/T356-protected-headers.sh                | 33 +++++++++++++++
 .../protected-with-legacy-display.eml         | 40 +++++++++++++++++++
 2 files changed, 73 insertions(+)
 create mode 100644 test/corpora/protected-headers/protected-with-legacy-display.eml

diff --git a/test/T356-protected-headers.sh b/test/T356-protected-headers.sh
index 4af018f3..43dfffe6 100755
--- a/test/T356-protected-headers.sh
+++ b/test/T356-protected-headers.sh
@@ -136,4 +136,37 @@ id:nested-rfc822-message@crypto.notmuchmail.org
+test_begin_subtest "when rendering protected headers, avoid rendering legacy-display part"
+output=$(notmuch show --format=json id:protected-with-legacy-display@crypto.notmuchmail.org)
+test_json_nodes <<<"$output" \
+                'subject:[0][0][0]["headers"]["Subject"]="Interrupting Cow"' \
+                'no_legacy_display:[0][0][0]["body"][0]["content"][1]["content-type"]="text/plain"'
+test_begin_subtest "when replying, avoid rendering legacy-display part"
+output=$(notmuch reply --format=json id:protected-with-legacy-display@crypto.notmuchmail.org)
+test_json_nodes <<<"$output" \
+                'no_legacy_display:["original"]["body"][0]["content"][1]["content-type"]="text/plain"'
+test_begin_subtest "do not treat legacy-display part as body when indexing"
+output=$(notmuch search --output=messages body:interrupting)
+test_expect_equal "$output" ''
+test_begin_subtest "identify message that had a legacy display part skipped during indexing"
+output=$(notmuch search --output=messages property:index.repaired=skip-protected-headers-legacy-display)
+test_expect_equal "$output" id:protected-with-legacy-display@crypto.notmuchmail.org
+# TODO: test that a part that looks like a legacy-display in
+# multipart/signed, but not encrypted, is indexed and not stripped.
+# TODO: test that a legacy-display in a decrypted subpart (not in the
+# cryptographic payload) is indexed and not stripped.
+# TODO: test that a legacy-display inside multiple MIME layers that
+# include an encryption layer (e.g. multipart/encrypted around
+# multipart/signed) is stripped and not indexed.
diff --git a/test/corpora/protected-headers/protected-with-legacy-display.eml b/test/corpora/protected-headers/protected-with-legacy-display.eml
new file mode 100644
index 00000000..8c5dd251
--- /dev/null
+++ b/test/corpora/protected-headers/protected-with-legacy-display.eml
@@ -0,0 +1,40 @@
+From: test_suite@notmuchmail.org
+To: test_suite@notmuchmail.org
+Subject: Subject Unavailable
+Date: Sat, 01 Jan 2000 12:00:00 +0000
+Message-Id: <protected-with-legacy-display@crypto.notmuchmail.org>
+MIME-Version: 1.0
+Content-Type: multipart/encrypted; boundary="=-=-=";
+	protocol="application/pgp-encrypted"
+Content-Type: application/pgp-encrypted
+Version: 1
+Content-Type: application/octet-stream
+-----END PGP MESSAGE-----

notmuch mailing list