PGPENCRYPT, PGPKEY, SMIMEENCRYPT, and SMIMEOPAQUE are all combination
flags, with multiple bits set.  In a few places these flags were
bitwise-and'ed incorrectly: expecting a non-zero result to indicate
all the bits in the flag were set.  Change those to explicitly compare
the result against the original flag.

The $postpone_encrypt and $(pgp/smime)_self_encrypt configuration
variables have created a somewhat messier situation for users.  Many
of them now have to specify their keys across multiple configuration
variables.

(Trac) Ticket #3983 had a reasonable request: "if my encrypt and
signing keys are the same, why can't I just specify my key once in my
.muttrc?"

The problem currently is that $smime_default_key and $pgp_sign_as are
both used to specify signing keys, and are set by the "sign (a)s"
security menu choice.  So we can't store encryption keys there because
some users have separate sign-only capability keys.

Create $pgp_default_key to store the default encryption key.  Change
signing to use $pgp_default_key, unless overridden by $pgp_sign_as.
The pgp "sign (a)s" will continue setting $pgp_sign_as.

Create $smime_sign_as.  Change signing to use $smime_default_key
unless overridden by $smime_sign_as.  Change s/mime "sign (a)s" menu
to set $smime_sign_as instead.

Change $postpone_encrypt and $(pgp/smime)_self_encrypt to use
$(pgp/smime)_default_key by default.

Mark $(pgp/smime)_self_encrypt_as deprecated.  They are now aliases
for the $(pgp/smime)_default_key config vars.

Change $(pgp/smime)_self_encrypt default to set.

The intent is that most users now need only set
$(pgp/smime)_default_key.  If they have a sign-only key, or have
separate signing and encryption keys, they can put that in
$(pgp/smime)_sign_as.  This also enables to default self_encrypt on
and solve a very common request.

Thanks to Michele Marcionelli and Vincent Lefèvre for gently pushing
me towards a solution.

Add $mime_type_query_command to specify a command to run if the
attachment extension is not in the mime.types file.

Add $mime_type_query_first to allow the query command to be run before
the mime.types lookup.

Type text/plain should not contain any null characters.  Slightly
improve the type guesser by forcing an attachment with any null
characters to be application/octet-stream.

Note the type guesser could use much more improvement, but this is an
easy and obvious fix.

Adding a mutt_endwin() seemed like a clean solution to allowing
ncurses pinentry for $sendmail, but it leaves other users watching a
blank screen.  This change is extremely likely to generate a large
number of complaints and bug reports.  So instead, force a hard
refresh afterwards.

This is to support invoking a program such as msmtp that can use gpg
to decrypt a password.  The ncurses pinentry can corrupt the screen
unless we exit curses.

Changeset fa1192803257 converted all exec calls to use mutt_envlist().
Unfortunately, the call in sendlib.c, execvpe(), is a glibc extension.

Convert back to execvp() for now, to fix the build on MacOS.

The documentation implies that all children processes will be affected
by the setenv command, so convert all the exec calls to use
mutt_envlist().

The setenv("GPG_TTY") call is no longer needed so remove it.

With that removed, there are no other setenv calls in mutt, so remove
the autoconf check and replacement function.

If the delimiter exists, additional sendmail flags will be inserted
before the delimiter.  Any arguments after the delimiter will be
preserved as recipients.

mutt_message_to_7bit() and transform_to_7bit() were overwriting
d_filename without freeing the previous value.

mutt_free_body() was not freeing the d_filename pointer.

mutt_get_content_info() was directly setting charset without freeing the previous value.

mutt_free_body() was not freeing the charset.

The mx_close_message is one of the few mx_* functions that don't have a
context as parameter. To make them more consistent, pass the context.

Changeset 23334e967dd7 created a workaround for a namespace conflict
with Solaris and derivatives.  After some discussion, the team decided
it would be best to move away from using the "M_" prefix for macros.

This patch was automatically generated by running:
  perl -wpi -e 's/\bM_(\w+)\b/MUTT_$1/g' `find . -name '*.[ch]' -print`
with the exception that sys_socket.h was exempted.  (That file will
be backed out subsequent to this commit.)

Thanks to Andras Salamon for supplying the perl script used to make
this change.

Most of these were just message update/clearing.

The Doctor reported triggering an assertion in convert_file_to(), due
to an unexpected errno from iconv().  According to the comments, the
assertions were only enabled for debugging and should have been turned
off.  We certainly don't want to abort mutt for this case, so just
disable them as the comment indicates.

This patch only updates existing copyright notices in the source
files, using commit dates since the last copyright update in commits
e3af935cdb1a and f8fd60d8d3f2.

Add a notice to the COPYRIGHT file to refer to our mercurial
repository for the full commit history.

Add myself to the COPYRIGHT file and smime_keys.pl file.

This is patch 1 of 4 implementing support for SMTPUTF8 (RFC 6531).

Change mutt_idna.c to be always compiled.  Remove the stub functions in
mutt_idna.h.  Instead, put #ifdefs around the idna function calls.  The
conversion functions will be fixed up in the next patch.

Rename the conversion functions to mutt_addrlist_to_intl() and
mutt_env_to_intl().  Rename the ADDRESS idna status bits to "intl"
status bits.

The output of mutt_local_tz() was being passed to abs().  Technically
the return type is time_t, but it represents a small value: the timezone
offset in seconds.  Add a safe explicit cast to int.

Change the txt parameter of mutt_make_help() to type const char *.
Typically all calls run the txt parameter through _(), which
accepts const char * and returns a char *. However, if NLS is not
enabled, _() is a noop, simply returning the parameter itself.  In
mutt_compile_help(), items[i].name is const char *, so it will generate
a warning when passed as the txt parameter of mutt_make_help().

On some systems, e.g. OS X, snprintf is defined as a macro.  One call
in hcache.c was embedding directives inside the snprintf call.  This is
apparently undefined behavior, so duplicate the call instead.

This allows oppenc to be enabled/disabled on a message level.  If
something initially enables encryption, such as crypt_autoencrypt or
crypt_replyencrypt, oppenc is turned off for the message.

Change the postpone/resume code to persist the oppenc bit.

Also change resend message to enable and invoke oppenc if the option is
set.

When a header has no value (*p == '\0' so l == 0), do not read outside
the buffer but print the newline anyway when a tag is already printed
(col != 0).

Caught by ASAN while opening a draft with no Subject.

This patch adds a mx_open_new_message() flag, M_SET_DRAFT.  It also adds
a MESSAGE->flags.draft flag.

mutt_write_fcc() passes the M_SET_DRAFT flag to mx_open_new_message(),
which then sets MESSAGE->flags.draft.  Then, imap_append_message() is
able to see this flag and so adds the \Draft flag.

The imap_append_message() function started to have a bit too many flags,
so this version of the patch separates out the flag generating code into
a simpler version.

This patch fixes CVE-2014-9116 in the stable branch.  It reverts
write_one_header() to the pre [f251d523ca5a] code for skipping
whitespace.

Thanks to Antonio Radici and Tomas Hoger for their analysis and patches
to mutt, which this patch is based off of.

closes #3635

add new inline funtion skip_email_wsp() to be used in lieu of SKIPWS() when parsing ASCII protocols rather than user input.

change use of SKIPWS() to skip_email_wsp() in places where it is likely to be a problem.

pointer > end pointer, as a result of using SKIPWS().  This
occurred because the header field had an empty body, so the
trailing newline was skipped over.  The fix is to just skip over
ascii space and htab.

closes #3609

Many structs used in mutt are actually constant but are defined
without the 'const' keyword. This can slow initialization (slightly)
in some environments due to extra copying and increases the amount of
writable RAM required at run-time, which can be significant on non-MMU
systems. Using const can also increase the opportunities for compiler
optimization.

The attached patch marks many such structures as const. On my test x86
build, this reduces the size of .data by over 50%.

add DISPNONE to mean no preferred Content-Disposition; make pgp-signature parts DISPNONE for multipart/signed

Closes #3385.

From Rocco's patch queue.