This is *not* based on autoconf. In fact, this doesn't actually
configure anything, (one can compile notmuch directly with just
"make" without running configure if the dependencies are all
satisfied).
The only thing that this configure script does is to check for the
presence of the various dependencies and provide some guidance to
the user if they are not all available.
I had noticed several times earlier that having a talloc context
passed in would make things more convenient. I'm not exercising
that convenience yet, but the context is there now, (and there's
one fewer item on our TODO list).
These have keybindings of '+', '-', and 'a'. The bug they have so
far is lack of visual feedback for their effect, and lack of undo.
(Also the fact that adding or removing a single tag for a thread
takes way too long--but that's as a Xapian issue as discussed here:
replace_document should make minimal changes to database file
http://trac.xapian.org/ticket/250
)
We now get the point staying right at the top where we want it.
We also don't get any extraneous noise about "Process notmuch
completed" or anything like that. Just the output in a read-only
buffer.
Compilation mode does a bunch of things that we don't want. Instead
of trying to tear it down to what we want, let's start at the other
end and build up only things that we really want.
I'm using that file as my reference here, so I'm likely to end up
copying some code here or there. Might as well be safe and just
copy the copyright statement.
Doesn't really do anything so far other than mark the buffer read-
only. This does have the benefit of giving us our own name rather
than "Compilation" for the mode.
The timestamp stuff we'll want to do soon, since it's a database
change, (though not a major one---at worst a handful of stale
timestamp documents would be left in the database).
We were aware of this bug when we wrote the function, (that a date
six days in the past would be treated as the "Friday" or as the
"Oct. 23" case depending on whether its time was before or after
the current time today). We thought it wouldn't be a problem, but
in practice it is. In scanning search results with this output,
the transition between formats makes it look like a day boundary,
(so it would be easy to mistakenly think "Oct. 23" is Thursday).
Fix this to avoid confusion, (still being careful to never print
"Thursday" for a date 7 days in the past when today is Thursday).
We're using a delimiter syntax that Keith is optimistic about
being able to easily parse in emacs. Note: We're not escaping
any occurrence of the delimiters in the message yet, so we'll
need to fix that.
With the recent addition of full-text indexing, printing only once per
1000 files just isn't often enough. The new timer-based approach will
be reliable regardless of the speed of adding message.
Our old notmuch-index-message.cc code had this, but I originally
left it out when adding indexing back in. I was concerned primarily
with mistakenly detecting signature markers and omitting important
text, (for example, I often do long lines of "----" as section
separators).
But now I see that there's a performance benefit to skippint the
quotations, (about 120 files/sec. instead of 95 files/sec.). I mitigated
the bogus signature checking by recognizing nothing other than the
all-time classic "-- ".
This avoids us wasting a bunch of time doing an expensive SHA-1 over a large
file only to discover later that it doesn't even *look* like an email message.
We put these is as a separate term so that they can be extracted.
We don't actually need this for searching, since typing an email
address in as a search term will already trigger a phrase search
that does exactly what's wanted.
This is based on the old notmuch-index-message.cc from early in
the history of notmuch, but considerably cleaned up now that
we have some experience with Xapian and know just what we want
to index, (rather than just blindly trying to index exactly
what sup does).
This does slow down notmuch_database_add_message a *lot*, but I've
got some ideas for getting some time back.
The original documentation of implicit AND is what we want, but
Xapian doesn't actually let us get that today. So be honest about
what the user can actually expect. And let's hope the Xapian
wizards give us the feature we want soon:
http://trac.xapian.org/ticket/402
"notmuch tag" is implemented now and seems to work great (and fast).
As for the race condition, as noted in the description we're removing
it's not exposed directly in the API, but only in a client that
allows for looping over search results and removing the inbox tag
from all of them. But then, that's exactly what the "notmuch tag"
command does. So, as discussed, we've now documented that command
to highlight the issue. Problem resolved, (as well as we can).
Putting all of our documentation into a single help message was getting
a bit unwieldy. Now, the simple output of "notmuch help" is a reasonable
reminder and a quick reference. Then we now support a new syntax of:
"notmuch help <command>" for the more detailed help messages.
This gives us freedom to put more detailed caveats, etc. into some
sub-commands without worrying about the usage statement getting too
long.
We were nicely reporting the lock-aquisition failure, but then marching
along trying to use the database object and just crashing badly.
So don't do that.
It's convenient to be able to do things like:
notmuch tag -inbox thread:<thread-id>
(even though this can run into a race condition as noted in TODO--the fix
for the race is simply to not run "notmuch new" between reading a thread
with the (not yet existent) "notmuch show" and removing its inbox tag
with a command like the above). So we now allow such a thing.
This uses the same search functionality as "notmuch search" so
it should be quite powerful. And this global search might be
quick enough to be used for "automatic" adding of tags to new
messages.
Of course, this will all be a lot more useful when we can search
for actual text of messages and not just tags.
The recent, disastrous failure of "notmuch new" would have been
avoided with this change. The new_command function was basically
assuming that it would only get a message object on success so
wasn't destroying the message in the other cases.
This would have helped with the recent bug causing "notmuch new"
to not record any results in the database. I'm not sure why
the explicit flush would be required, (shouldn't the destructor
always ensure that things flush?), but perhaps some outstanding
references from the leak prevented that.
In any case, an explicit flush on close() seems to make sense.
I'm trying to stick to a habit of fixing previously-introduced bugs
on side branches off of the commit that introduced the bug. The
idea here is to make it easy to find the commits to cherry pick
if bisecting in the future lands on one of the broken commits.
We were incorrectly only destroying messages in the case of
successful addition to the database, and not in other cases,
(such as failure due to FILE_NOT_EMAIL).
I'm still not entirely sure why this was performing abysmally, (as in
making an operation that should take a small fraction of a second take
10 seconds), nor why it was causing the database to entirely fail to
get new results.
But fortunately, this all seems to work now.
