Gnus asking for key twice

[Adam]

Whenever I send a signed mail which has a Gcc'ed header pointing to my
archive group, I am asked to select the sign key twice - once before
connecting to my smtp server, and once after actually sending the mail.
mm-sign-option is set to guided. When there is no Gcc header, I am only
asked once.

Is this a bug or a feature?

[Adam]

Post by Adam
Whenever I send a signed mail which has a Gcc'ed header pointing to my
archive group, I am asked to select the sign key twice - once before
connecting to my smtp server, and once after actually sending the mail.
mm-sign-option is set to guided. When there is no Gcc header, I am only
asked once.
Is this a bug or a feature?

I investigated. gnus-inews-do-gcc is calling
message-encode-message-body, which internally triggers the key selection
dialogue. This looks like a real design flaw to me. Gnus should encode
the message only once, and preserve the result.

[Adam]

Post by Adam

Post by Adam
Whenever I send a signed mail which has a Gcc'ed header pointing to my
archive group, I am asked to select the sign key twice - once before
connecting to my smtp server, and once after actually sending the mail.
mm-sign-option is set to guided. When there is no Gcc header, I am only
asked once.
Is this a bug or a feature?

I investigated. gnus-inews-do-gcc is calling
message-encode-message-body, which internally triggers the key
selection dialogue. This looks like a real design flaw to me. Gnus
should encode the message only once, and preserve the result.

The easiest solution is to add a new hook. It should be called in
messages-send-mail right after the message has been dispatched to the
server, with the current buffer being the actual message buffer that has
been sent. Gnus then should add gnus-inews-do-gcc to this hook, and use
the processed buffer to save.

Do I sound reasonable? Whats a good name for such a new hook?

[Katsumi Yamaoka]

Post by Adam

Post by Adam

Post by Adam
Whenever I send a signed mail which has a Gcc'ed header pointing to my
archive group, I am asked to select the sign key twice - once before
connecting to my smtp server, and once after actually sending the mail.
mm-sign-option is set to guided. When there is no Gcc header, I am only
asked once.
Is this a bug or a feature?

I investigated. gnus-inews-do-gcc is calling
message-encode-message-body, which internally triggers the key
selection dialogue. This looks like a real design flaw to me. Gnus
should encode the message only once, and preserve the result.

The easiest solution is to add a new hook. It should be called in
messages-send-mail right after the message has been dispatched to the
server, with the current buffer being the actual message buffer that has
been sent. Gnus then should add gnus-inews-do-gcc to this hook, and use
the processed buffer to save.
Do I sound reasonable? Whats a good name for such a new hook?

A sent message and Gcc'd one are not exactly the same. AFAICT,
the Gcc header remains in a Gcc copy but is deleted in the one
sent. Maybe there are other differences besides I cannot recall.
So, we will have to verify all the message sending functions and
situations in order to make such a change, I think.

How about making the expiry time of the passphrase cache longer?
The `mml-secure-passphrase-cache-expiry' controls it; the default
value might be too short. If you want to try changing it briefly
where mml2015.elc has already been loaded, you'll need to change
`mml2015-passphrase-cache-expiry' instead (it inherits the value
of `mml-secure-passphrase-cache-expiry' as the default value).

[Katsumi Yamaoka]

AFAICT, the Gcc header remains in a Gcc copy but is deleted in the
one sent. ^^^^^^^^^^

Oops, I meant the X-Draft-From header and the like.

[Adam]

Post by Katsumi Yamaoka
How about making the expiry time of the passphrase cache longer?
The `mml-secure-passphrase-cache-expiry' controls it; the default
value might be too short. If you want to try changing it briefly
where mml2015.elc has already been loaded, you'll need to change
`mml2015-passphrase-cache-expiry' instead (it inherits the value
of `mml-secure-passphrase-cache-expiry' as the default value).

The passphrase is not the problem. In fact I do not protect my keys
with passphrases. It is the actual key (say the key that should be used
to sign the message) selection dialogue that pops up twice.

[Katsumi Yamaoka]

Post by Adam
The passphrase is not the problem. In fact I do not protect my keys
with passphrases. It is the actual key (say the key that should be used
to sign the message) selection dialogue that pops up twice.

I overlooked you wrote `mm-sign-option is set to guided', sorry.
That you are prompted twice seems to be due to a bug of
`mml2015-epg-sign'. The encrypting function might have it, too.
There are two issues in the following section:

(or (message-options-get 'mml2015-epg-signers)
(message-options-set 'mml2015-epg-signers ...))

The second line tries to save a signer to the `message-options'
variable and the first line gets it back in the next time.
However, `message-options' is always buffer-local and those are
done in a temporary buffer. So, this save-and-back plan goes
wrong.

In addition, `epa-select-keys', that passes a signer to the second
line from the selection menu, always returns nil as far as I can
observe.

It seems hard to fix.

[Katsumi Yamaoka]

Post by Katsumi Yamaoka
That you are prompted twice seems to be due to a bug of
`mml2015-epg-sign'. The encrypting function might have it, too.
(or (message-options-get 'mml2015-epg-signers)
(message-options-set 'mml2015-epg-signers ...))
The second line tries to save a signer to the `message-options'
variable and the first line gets it back in the next time.
However, `message-options' is always buffer-local and those are
done in a temporary buffer. So, this save-and-back plan goes
wrong.

Fixed. The value of `message-options' is shared between a draft
buffer and temporary working buffers now. So, a signer you once
select is kept in it and you should never be prompted for a signer
twice even if you set `mm-sign-option' to `guided'.

Post by Katsumi Yamaoka
In addition, `epa-select-keys', that passes a signer to the second
line from the selection menu, always returns nil as far as I can
observe.

I should have used the `m' key, not the mouse-2 button, to select
one.

[Adam]

Post by Katsumi Yamaoka

Post by Katsumi Yamaoka
That you are prompted twice seems to be due to a bug of
`mml2015-epg-sign'. The encrypting function might have it, too.
(or (message-options-get 'mml2015-epg-signers)
(message-options-set 'mml2015-epg-signers ...))
The second line tries to save a signer to the `message-options'
variable and the first line gets it back in the next time.
However, `message-options' is always buffer-local and those are
done in a temporary buffer. So, this save-and-back plan goes
wrong.

Fixed. The value of `message-options' is shared between a draft
buffer and temporary working buffers now. So, a signer you once
select is kept in it and you should never be prompted for a signer
twice even if you set `mm-sign-option' to `guided'.

Post by Katsumi Yamaoka
In addition, `epa-select-keys', that passes a signer to the second
line from the selection menu, always returns nil as far as I can
observe.

I should have used the `m' key, not the mouse-2 button, to select
one.

Thank you