b4 prep and --manual-reroll

Commit Message

Dmitry Torokhov Nov. 4, 2022, 5 a.m. UTC

Hi Konstantin,

Somehow "b4 send -o dir" followed by "git send-email ..." managed to
produce quoted-printable cover letter message (see
https://lore.kernel.org/all/[e-mail]/)
and that confused "b4 prep --manual-reroll ..." as this produced a tag
that had "=0D" as line ends. I was able to work around it with the
patch below (not sure if this is right thing to do)...

Also, doing "b4 send --resend" with this series did produce double
signature on the cover letter, even though I see in the code it tries to
remove it.

Thanks.

Comments

Konstantin Ryabitsev Nov. 4, 2022, 5:13 p.m. UTC | #1

On Thu, Nov 03, 2022 at 10:00:07PM -0700, Dmitry Torokhov wrote:
> Hi Konstantin,
> 
> Somehow "b4 send -o dir" followed by "git send-email ..." managed to
> produce quoted-printable cover letter message (see
> https://lore.kernel.org/all/[e-mail]/)
> and that confused "b4 prep --manual-reroll ..." as this produced a tag
> that had "=0D" as line ends. I was able to work around it with the
> patch below (not sure if this is right thing to do)...

Ah, yes, quoted-printable is the bane of my existence. Thank you for reporting
this. I think a better way forward is to get away from get_payload(), which is
a legacy method that I kept around when we cared about python3.6
compatibility. I'll comb the codebase for a better solution.

> Also, doing "b4 send --resend" with this series did produce double
> signature on the cover letter, even though I see in the code it tries to
> remove it.

I'll poke at it -- thanks for testing that out. I'm pretty sure you're the
first person to actually use it. :)

-K

Rob Herring (Arm) Nov. 4, 2022, 5:47 p.m. UTC | #2

On Fri, Nov 4, 2022 at 12:14 PM Konstantin Ryabitsev
<[e-mail]> wrote:
>
> On Thu, Nov 03, 2022 at 10:00:07PM -0700, Dmitry Torokhov wrote:
> > Hi Konstantin,
> >
> > Somehow "b4 send -o dir" followed by "git send-email ..." managed to
> > produce quoted-printable cover letter message (see
> > https://lore.kernel.org/all/[e-mail]/)
> > and that confused "b4 prep --manual-reroll ..." as this produced a tag
> > that had "=0D" as line ends. I was able to work around it with the
> > patch below (not sure if this is right thing to do)...
>
> Ah, yes, quoted-printable is the bane of my existence. Thank you for reporting
> this. I think a better way forward is to get away from get_payload(), which is
> a legacy method that I kept around when we cared about python3.6
> compatibility. I'll comb the codebase for a better solution.

FWIW, this is what I use:

body = orig_msg.get_body(preferencelist='plain').get_content(errors='strict')


For sending replies, python prefers base64 if any encoding is needed,
so I do this:

    msg = email.message.EmailMessage(policy=email.policy.default)
    charset = orig_msg.get_content_charset()
    if charset:
        msg.set_content(body, charset=charset, cte='8bit')
    else:
        msg.set_content(body, charset='ascii', cte='7bit')


Rob

Patch

diff --git a/b4/ez.py b/b4/ez.py
index 4fcc988..8fb39d2 100644
--- a/b4/ez.py
+++ b/b4/ez.py
@@ -1350,7 +1350,7 @@  def get_sent_tagname(branch: str, tagprefix: str, revstr: Union[str, int]) -> Tu
 def reroll(mybranch: str, cover_msg: email.message.Message, tagprefix: str = SENT_TAG_PREFIX):
     # Prepare annotated tag body from the cover letter
     lsubject = b4.LoreSubject(cover_msg.get('subject'))
-    cbody = cover_msg.get_payload()
+    cbody = cover_msg.get_payload(decode=True).decode()
     # Remove signature
     chunks = cbody.rsplit('\n-- \n')
     if len(chunks) > 1: