Reviewed-by: Carl Worth <cworth@cworth.org>
Keith wrote all the code here against notmuch before notmuch.c was
split up into multiple files. So I've pushed the code around in
various ways to match the new code structure, but have generally tried
to avoid making any changes to the behavior of the code.
I did fix one bug---a missing call to g_mime_stream_file_set_owner in
show_part which would cause "notmuch show" to go off into the weeds
when trying to show multiple messages, (since the first stream would
fclose stdout).
Now that the client sources are alone here in their own directory,
(with all the library sources down inside the lib directory), we can
break the client up into multiple files without mixing the files up.
The hope is that these smaller files will be easier to manage and
maintain.
Previously, the top-level Makefile was explicitly adding -I./lib to
the compiler flags. However, that's something that's much better done
from within the Makefile.local fragment within the lib directory
itself.
We were previously using separate CFLAGS and NOTMUCH_CFLAGS variables
in an attempt to allow the user to specify CFLAGS on the command-line.
However, that's just a lot of extra noise in the Makefile when we can
instead let the user specify what is desired for CFLAGS and then use
an override to append the things we require. So our Makefile is much
neater now.
I saw this recommendation in the implementation notes for "Recursive
Make Considered Harmful" and then the further recommendation for
implementing the idea in the GNU make manual.
The idea is that if any of the files change then we need to regenerate
the dependency file before we regenerate any targets.
The approach from the GNU make manual is simpler in that it just uses
a sed script to fix up the output of an extra invocation of the
compiler, (as opposed to the approach in the implementation notes from
the paper's author which use a wrapper script for the compiler that's
always invoked rather than the compiler itself).
The idea here is that every Makefile at each lower level will be an
identical, tiny file that simply defers to a top-level make.
Meanwhile, the Makefile.local file at each level is a Makefile snippet
to be included at the top-level into a large, flat Makefile. As such,
it needs to define its rules with the entire relative directory to
each file, (typically in $(dir)). The local files can also append to
variables such as SRCS and CLEAN for files to be analyzed for
dependencies and to be cleaned.
Instead of the old name of Makefile.dep. The idea being that the
user really doesn't need to see this by default, (and if debugging
the Makefile, the rules will make the name obvious).
Previously, when selecting a thread to view from the search buffer, we
would advance the point by one line before showing the thread, (so
that it would be ready to show the next thread once the user was done
with the current thread). This was annoying when the user temporarily
exited the thread view, (because the "wrong" thread was then selected
in the search view).
We get a more consistent experience by waiting to advance until the
user has finished viewing one thread and is ready to view the next.
This crash was introduced sometime recently, as previously things
worked fine when notmuch detected that a file is not an email.
We're definitely overdue for that test suite.
A recent "notmuch restore" command took *forever* for me. Obviously,
we need to fix the underlying performance bug in Xapian, but in the
meantime, a progress indicator would help.
I recently added a print of the subject line for use as part of a
two-line summary in the emacs client. But of course, the subject was
already being printed on the next line. So I didn't really need to add
anything, I could have just stopped hiding what was already
printed. Anyway, we now avoid printing it twice in a row.
I'd fixed this earlier when I had a private copy of GMime's
date-parsing code, but I lost the fix when I recently switched to
calling the GMime function.
The recent change of the hidden thread-ID syntax caused this message
to instead be replaced with a cryptic "search failed" error and an
internal regular expression. Put our nice message back.
This is in place now citations and signatures. We'll still need to
add something else for hidden messages (those that are already
read and hidden away).
The more general command is more consistent, and more useful.
We also fix "notmuch search" to output copy-and-pasteable search terms
for the thread with "thread:" prepended already. Similarly, the
message-ID in the output of "notmuch show" is also now printed as a
valid search term, ("id:<message-id>" rather than "ID: <message-id>").
Naturally, the emacs code is also changed to track these changes.
The magic space bar is nice, but sometimes there's a message with a
long attachment that I just want to skip, but still consider the
message marked as read.
I had implemented this intentionally originally, thinking that it
would be important to see the last message scroll all the way off
screen before the next press of the magic space bar would go and
archive away the whole thread.
But in practice, that just turns out to be annoying, (especially for a
long sequence of single-message threads where the space bar has to be
pressed twice for every one). It's actually quite easy to know if it's
"safe" to press the space bar expecting just a scroll instead of an
archive by simply looking down and seeing if the current window is
full.
And as for the total lack of undo with all of this, I'm getting by by
simply using x to get back to the search view, and then going back
into the thread of interest.
More magic for the magic space bar: If a thread is entirely open,
(such as when viewing an old thread where every message is read), the
space bar now visits each message in turn (rather than skipping all of
the unread messages).
Otherwise known as "Backspace" on keyboards in the real, (rather than
emacs), world. This will go by screenfuls for long messages, and
message by message for short messages. So it does the reverse of the
magic space bar, (but without reversing any tag-changing magic that
the magic space bar might have done).
Chances are, a signature above a certain threshold isn't just a
signature, (for example, it could be an encrypted messages tacked onto
the end of the file, or could be any sort of PS.)
We add a new variable, notmuch-show-signature-lines-max that can be
used to configure the threshold, (set to 6 by default for now).
We were inadvertently calling g_object_unref on a wild pointer leading
to the following error message:
GLib-GObject-CRITICAL **: g_object_unref: assertion
`G_IS_OBJECT (object)' failed
Now, why glib doesn't abort on critical errors, I'll never understand.
I noticed that these functions would sometimes leave point on an
invisible character[*]. The problem would be that point would appear
to be on a particular message, but adding or removing a tag would
actually add/remove a tag from the *previous* message.
Fix the C-n and C-p keybindings at least to call the underlying
command and then advance to a visible character. We set this-command
in our overrides so that the temporary-goal-column feature still
works.
[*] The documentation says that command loop is supposed to move point
outside of any invisible region when a command exits. But apparently
not.
Clearly some recent code was very fragile, which I noticed in that the
space bar would no longer scroll a long message if it was the only
message in a thread.
This resulted in a lot of churn, but hopefully things are more robust
now, (for example by using new predicates like
notmuch-show-last-message-p rather than doing heuristics based on
(eobp) or (window-end)).
As usual, the presence of invisible characters complicates the task of
making this stuff robust.
With the recent change of showing the first unread message, we would
scroll down to the end of the buffer if all messages were already
read. This would confusingly show nothing visible in the window.
Instead, detect this case and move to the beginning of the buffer.
The magic of the space bar is all about unread messages, so there's no
reason for it to advance to messages that have already been read.
Similarly, we now remove any magic from (n)ext so that it simply
advances to the next message without marking anything read, (which
makes it symmetrical with (p)revious).
This allows for pleasant termination of the "show next thread" magic
in notmuch-show mode. Now, it will terminate and show the
notmuch-search results rather than continually displaying the last
thread over and over.
This is implemented by stashing away the parent notmuch-search buffer
into a variable within the notmuch-show buffer. Then, when magic space
bar triggers an archive of the current thread, it switches to the parent
search buffer and shows the next thread.
