Placing two semicolons on an otherwise empty line helps to logically
"connect" the surrounding "paragraphs", which in (only) some cases
makes sense.
Previously the three paragraphs of the permission statement were not
connected to each other like this, which is perfectly fine. However
the preceding "This file is not part of GNU Emacs." line was connected
to the first paragraph, which does not make sense considering that the
latter is not connected two the second paragraph, which it relates to
more.
Once those two semicolons are gone, it also makes sense to remove
those from the second line.
Commit edce635416 only loads it
dynamically when executing the function mu4e~view-gnus but as a
consequence many variables are undefined at compile time and, more
importantly, it is no longer possible to override the variable
gnus-display-mime-function with a let binding before calling
gnus-article-prepare-display
Disable some of the key-binding that don't quite work for the gnus-based
view. We'll probably want a gnus-view specific set of bindings, but at
least for now avoid the errors.
Previously both cl-lib.el and cl.el were used, now use only cl-lib.el.
Use lexical-binding where needed instead of requiring cl just for
`lexical-let`.
Replace some add-to-list with cl-pushnew as add-to-list is not
recommended in lisp program and anyway doesn't work properly with
lexical binding.
This is a bit of hack, but solves the problem; even when not _showing_
the attachments-header, construct the corresponding map, so the
keybindings work.
Mark some variables as use permanent-local so they'll survive a
major-mode change. Activate mu4e-view later, so hooks have some more
material to work with.
Since cd649efb6b, opening an unread message first does a proc-move,
then proc-view.
Reason is that while we get the (:update ... ) from the move, that only
contains a skeleton message; we need the full view get images etc. This
means that we render the message _twice_.
Here we change add a flag for move to _not_ send the (:update ..), so
only the (:view ...) will trigger rendering of the message.
When we render URLs like https://gnu.org/[1] right-clicking on them in
e.g. GNOME Terminal will also copy the "[1]" as well as the URL,
inserting zero-width-space[1] between the two avoids this.
I know about "g" (mu4e-view-go-to-url), but sometimes I want to open a
URL in a different browser, or copy it into a non-Emacs program. This
makes that easier. I think this improves the UI at a very trivial cost
to users that don't care about this use-case.
I could make this configurable, but unless someone vehemently objects
to this I don't see the point of not just making it the default.
In GNOME Terminal a ZWS is rendered simply as a space, and
copy/pasting works as expected, but in Emacs's GTK GUI there's no
space between the two.
This was initially a plain ASCII space character, but djcb preferred a
ZWP, and this works as well.
1. https://en.wikipedia.org/wiki/Zero-width_space
Define the view-mode (gnus or internal) just before it's needed. This
ensures that (when in gnus mode) it's really a gnus-article-mode
derivative, so more of the gnus specifics work.
When we mark a message as read, we get an (:update ... ) with the
marked-as-read message (ie., moved from new/ to cur/).
The (:update ...) however does _not_ include extracted images etc.; so
images in unread message would not be visible at that time. To fix this,
to another (:view ...) and extract the images etc.
Thanks to tangxinfa for pointing this out.
Add a defcustom `mu4e-view-use-gnus`, which, when `t', use Gnus'
article-mode for displaying messages instead of mu4e's built-in mode.
This is experimental, and a view messages do not display
correctly yet (base64-encoded bodies).
before text insertion and run finally the hook at end.
This ensure local variables are not killed after being set.
* mu4e/mu4e-view.el (mu4e-view): Do it.