[PATCH 02/11] test: New tests for Emacs charset handling

Thu Apr 24 23:18:35 PDT 2014

On Thu, 24 Apr 2014, Austin Clements <amdragon at MIT.EDU> wrote:
> Quoth Mark Walters on Apr 24 at  3:38 pm:
>> 
>> On Mon, 21 Apr 2014, Austin Clements <amdragon at MIT.EDU> wrote:
>> > The test of viewing 8bit messages is known-broken.  The rest pass, but
>> > for very fragile reasons.  The next several commits will fix the
>> > known-broken test and make our charset handling robust.
>> 
>> Hi 
>> 
>> On one of my systems one of these (non-broken) tests fails. I am not
>> sure whether I messed up my emacs/environment when doing stuff remotely
>> recently so it could just be my system
>> 
>> 
>> > ---
>> >  test/T455-emacs-charsets.sh | 141 ++++++++++++++++++++++++++++++++++++++++++++
>> >  test/test-lib.el            |   4 +-
>> >  2 files changed, 144 insertions(+), 1 deletion(-)
>> >  create mode 100755 test/T455-emacs-charsets.sh
>> >
>> > diff --git a/test/T455-emacs-charsets.sh b/test/T455-emacs-charsets.sh
>> > new file mode 100755
>> > index 0000000..a42a1d2
>> > --- /dev/null
>> > +++ b/test/T455-emacs-charsets.sh
>> > @@ -0,0 +1,141 @@
>> > +#!/usr/bin/env bash
>> > +
>> > +test_description="emacs notmuch-show charset handling"
>> > +. ./test-lib.sh
>> > +
>> > +
>> > +UTF8_YEN=$'\xef\xbf\xa5'
>> > +BIG5_YEN=$'\xa2\x44'
>> > +
>> > +# Add four messages with unusual encoding requirements:
>> > +#
>> > +# 1) text/plain in quoted-printable big5
>> > +generate_message \
>> > +    [id]=test-plain at example.com \
>> > +    '[content-type]="text/plain; charset=big5"' \
>> > +    '[content-transfer-encoding]=quoted-printable' \
>> > +    '[body]="Yen: =A2=44"'
>> > +
>> > +# 2) text/plain in 8bit big5
>> > +generate_message \
>> > +    [id]=test-plain-8bit at example.com \
>> > +    '[content-type]="text/plain; charset=big5"' \
>> > +    '[content-transfer-encoding]=8bit' \
>> > +    '[body]="Yen: '$BIG5_YEN'"'
>> > +
>> > +# 3) text/html in quoted-printable big5
>> > +generate_message \
>> > +    [id]=test-html at example.com \
>> > +    '[content-type]="text/html; charset=big5"' \
>> > +    '[content-transfer-encoding]=quoted-printable' \
>> > +    '[body]="<html><body>Yen: =A2=44</body></html>"'
>> > +
>> > +# 4) application/octet-stream in quoted-printable of big5 text
>> > +generate_message \
>> > +    [id]=test-binary at example.com \
>> > +    '[content-type]="application/octet-stream"' \
>> > +    '[content-transfer-encoding]=quoted-printable' \
>> > +    '[body]="Yen: =A2=44"'
>> > +
>> > +notmuch new > /dev/null
>> > +
>> > +# Test rendering
>> > +
>> > +test_begin_subtest "Text parts are decoded when rendering"
>> > +test_emacs '(notmuch-show "id:test-plain at example.com")
>> > +	    (test-visible-output "OUTPUT.raw")'
>> > +awk 'show {print} /^$/ {show=1}' < OUTPUT.raw > OUTPUT
>> > +cat <<EOF >EXPECTED
>> > +Yen: $UTF8_YEN
>> > +EOF
>> > +test_expect_equal_file OUTPUT EXPECTED
>> > +
>> > +test_begin_subtest "8bit text parts are decoded when rendering"
>> > +test_emacs '(notmuch-show "id:test-plain-8bit at example.com")
>> > +	    (test-visible-output "OUTPUT.raw")'
>> > +awk 'show {print} /^$/ {show=1}' < OUTPUT.raw > OUTPUT
>> > +cat <<EOF >EXPECTED
>> > +Yen: $UTF8_YEN
>> > +EOF
>> > +test_expect_equal_file OUTPUT EXPECTED
>> > +
>> > +test_begin_subtest "HTML parts are decoded when rendering"
>> > +test_emacs '(notmuch-show "id:test-html at example.com")
>> > +	    (test-visible-output "OUTPUT.raw")'
>> > +awk 'show {print} /^$/ {show=1}' < OUTPUT.raw > OUTPUT
>> > +cat <<EOF >EXPECTED
>> > +[ text/html ]
>> > +Yen: $UTF8_YEN
>> > +EOF
>> > +test_expect_equal_file OUTPUT EXPECTED
>> 
>> It's this test: I get an extra newline after the UFT8_YEN. 
>> 
>> This is with emacs 23.4.1 on a Debian wheezy(ish) system.
>> 
>> But as said it could be my fault: I think some odd things happened when
>> I tried to install emacs 24 and 23 simultaneously to help testing
>> things.
>
> I can reproduce this.  Tests that involve HTML rendering are always
> finicky like this since there are so many different HTML renderers.
> Maybe I could normalize the spacing?  sed '/^$/d;s/  */ /g' or so?

That sounds plausible. What output would you get if the test genuinely
fails? I guess the key thing is to try and make sure the test doesn't
pass in that case.

Best wishes

Mark