notmuch
3 years agolib: centralize query parsing, store results.
David Bremner [Sat, 25 Feb 2017 16:09:11 +0000 (12:09 -0400)]
lib: centralize query parsing, store results.

The main goal is to prepare the way for non-destructive (or at least
less destructive) exclude tag handling. It does this by having a
pre-parsed query available for further processing. This also allows us
to provide slightly more precise error messages.

3 years agolib: use delete[] to free buffer allocated using new[]
Jani Nikula [Sat, 4 Mar 2017 17:38:41 +0000 (19:38 +0200)]
lib: use delete[] to free buffer allocated using new[]

Fix warning caught by clang:

lib/regexp-fields.cc:41:2: warning: 'delete' applied to a pointer that was allocated
      with 'new[]'; did you mean 'delete[]'? [-Wmismatched-new-delete]
        delete buffer;
        ^
              []
lib/regexp-fields.cc:37:17: note: allocated with 'new[]' here
        char *buffer = new char[len];
                       ^

3 years agolib: add mid: as a synonym for id:
David Bremner [Mon, 27 Feb 2017 02:34:21 +0000 (22:34 -0400)]
lib: add mid: as a synonym for id:

mid: is the url scheme suggested by URL 2392. We also plan to
introduce more flexible searches for mid: than are possible with
id: (in order not to break assumptions about the special behaviour of
id:, e.g. identifying at most one message).

3 years agolib: regexp matching in 'subject' and 'from'
David Bremner [Mon, 27 Feb 2017 02:34:20 +0000 (22:34 -0400)]
lib: regexp matching in 'subject' and 'from'

the idea is that you can run

% notmuch search subject:/<your-favourite-regexp>/
% notmuch search from:/<your-favourite-regexp>/

or

% notmuch search subject:"your usual phrase search"
% notmuch search from:"usual phrase search"

This feature is only available with recent Xapian, specifically
support for field processors is needed.

It should work with bindings, since it extends the query parser.

This is easy to extend for other value slots, but currently the only
value slots are date, message_id, from, subject, and last_mod. Date is
already searchable;  message_id is left for a followup commit.

This was originally written by Austin Clements, and ported to Xapian
field processors (from Austin's custom query parser) by yours truly.

3 years agolib: create field processors from prefix table
David Bremner [Mon, 27 Feb 2017 02:34:19 +0000 (22:34 -0400)]
lib: create field processors from prefix table

This is a bit more code than hardcoding the two existing field
processors, but it should make it easy to add more.

3 years agobuild: use sha256sum instead of sha1sum to sign releases
David Bremner [Thu, 2 Mar 2017 00:44:47 +0000 (20:44 -0400)]
build: use sha256sum instead of sha1sum to sign releases

3 years agoemacs/desktop: update to use notmuch-emacs-mua and handle mailto
Jani Nikula [Sat, 25 Feb 2017 09:25:40 +0000 (11:25 +0200)]
emacs/desktop: update to use notmuch-emacs-mua and handle mailto

With the mailto: handling in notmuch-emacs-mua, we can update the
desktop file to advertize we can be set as the default application to
handle email. While at it, add GenericName and Comment to be more
informative.

With --hello, notmuch-emacs-mua will run (notmuch) if mailto: url is
not given.

3 years agocompletion: complete notmuch emacs-mua
Jani Nikula [Sat, 25 Feb 2017 09:25:39 +0000 (11:25 +0200)]
completion: complete notmuch emacs-mua

With subcommand handling for external commands we can easily complete
'notmuch emacs-mua' using the existing completion system.

3 years agoMerge branch 'release'
David Bremner [Wed, 1 Mar 2017 01:21:07 +0000 (21:21 -0400)]
Merge branch 'release'

Second gnugpg test suite fix

3 years agoupdate version to 0.23.7 0.23.7
David Bremner [Wed, 1 Mar 2017 00:49:24 +0000 (20:49 -0400)]
update version to 0.23.7

3 years agoNEWS: add news for 0.23.7
David Bremner [Wed, 1 Mar 2017 00:46:53 +0000 (20:46 -0400)]
NEWS: add news for 0.23.7

3 years agoadd changelog stanza for 0.23.7-1
David Bremner [Tue, 28 Feb 2017 00:07:52 +0000 (20:07 -0400)]
add changelog stanza for 0.23.7-1

3 years agonotmuch-config: ENOENT vs generic handling when file open fails.
Tomi Ollila [Sun, 26 Feb 2017 19:12:39 +0000 (21:12 +0200)]
notmuch-config: ENOENT vs generic handling when file open fails.

When opening configuration file fails, ENOENT (file not found) is
handled specially -- in setup missing file is ok (often expected),
and otherwise user can be informed to run notmuch setup.

In any other case the the reason is unknown, so there is no other
option but to print generic error message to stderr.

3 years agocli/config: don't try to open config file for 'notmuch help'
Jani Nikula [Sun, 26 Feb 2017 19:12:38 +0000 (21:12 +0200)]
cli/config: don't try to open config file for 'notmuch help'

The help command does not really need to try to open the config
file. So don't.

3 years agoMerge tag '0.23.6'
David Bremner [Tue, 28 Feb 2017 12:06:42 +0000 (08:06 -0400)]
Merge tag '0.23.6'

notmuch 0.23.6 release

3 years agocli/show: add content-disposition to structured output message parts
Jani Nikula [Sun, 26 Feb 2017 18:33:48 +0000 (20:33 +0200)]
cli/show: add content-disposition to structured output message parts

Help the clients decide how to display parts.

Test updates by Mark Walters <markwalters1009@gmail.com>.
One more test fix by db

3 years agocli/show: abstract get content disposition
Jani Nikula [Sun, 26 Feb 2017 18:33:47 +0000 (20:33 +0200)]
cli/show: abstract get content disposition

Reduce duplication in follow-up work. As a side effect, handle error
returns from g_mime_content_disposition_get_disposition() without
segfaulting.

3 years agotest: move GNUPGHOME to TEST_TMPDIR
David Bremner [Mon, 27 Feb 2017 23:53:56 +0000 (19:53 -0400)]
test: move GNUPGHOME to TEST_TMPDIR

We already use this directory for dtach sockets, so it makes sense to
put gnupg sockets there as well. There doesn't seem to be a clean way
to put a fully functional socket in a different location than
GNUPGHOME.

3 years agoRevert "configure: add test for gpgconf --create-socketdir"
David Bremner [Mon, 27 Feb 2017 22:27:49 +0000 (18:27 -0400)]
Revert "configure: add test for gpgconf --create-socketdir"

This reverts commit 12f7d4e61d30fb9a5c99e128ab1aab0bf069cdb4.

Since we're not using gpgconf anymore, drop the unneeded test in
configure.

3 years agoRevert "test: use gpgconf --create-socketdir if available"
David Bremner [Mon, 27 Feb 2017 22:26:47 +0000 (18:26 -0400)]
Revert "test: use gpgconf --create-socketdir if available"

This reverts commit e7b88e8b0a93ca83d807edc00e0c97af54c5b5f1.

It turns out that this does not work well in environments without a
running systemd (or some other provider of /run/user)

3 years agodebian: set upload date and suite 0.23.6
David Bremner [Mon, 27 Feb 2017 10:58:07 +0000 (06:58 -0400)]
debian: set upload date and suite

3 years agoNEWS: set release date
David Bremner [Mon, 27 Feb 2017 10:57:42 +0000 (06:57 -0400)]
NEWS: set release date

3 years agotest: make T640 compatible with old C compilers
David Bremner [Sun, 26 Feb 2017 12:16:45 +0000 (08:16 -0400)]
test: make T640 compatible with old C compilers

Apparently our test system does not use the same flags for compiling
tests as it does for compiling notmuch. Make the test compatible with
C89. Also remove one unused loop index.

3 years agocompletion: add bash completion for query: and property:
Jani Nikula [Sat, 17 Dec 2016 10:50:49 +0000 (12:50 +0200)]
completion: add bash completion for query: and property:

Add prefix completion for query and property, with value completion
for query. Apparently there's no way to get at the available
properties in the cli.

3 years agoemacs: show: stop display of application/* parts
Mark Walters [Sat, 28 Jan 2017 09:47:42 +0000 (09:47 +0000)]
emacs: show: stop display of application/* parts

Gnus seems to display application/zip and application/tar by
default. This doesn't seem desirable so we override it.

We only override if the user has not customized
mm-inline-override-types themselves.

3 years agotest: test format version difference between v2 and v3
Jani Nikula [Sat, 25 Feb 2017 13:31:32 +0000 (15:31 +0200)]
test: test format version difference between v2 and v3

Schemata v3 changed message filename field to a list. Test both
versions 2 and 3.

3 years agocli/show: list all filenames of a message in the formatted output
Jani Nikula [Sat, 25 Feb 2017 13:31:31 +0000 (15:31 +0200)]
cli/show: list all filenames of a message in the formatted output

Instead of just having the first filename for the message, list all
duplicate filenames of the message as a list in the formatted
outputs. This bumps the format version to 3.

3 years agoemacs: use (system-name) instead of system-name
Jani Nikula [Sun, 26 Feb 2017 09:28:30 +0000 (11:28 +0200)]
emacs: use (system-name) instead of system-name

Fix the deprecation warning:

In notmuch-maildir-fcc-make-uniq-maildir-id:
emacs/notmuch-maildir-fcc.el:279:53:Warning: ‘system-name’ is an obsolete
    variable (as of 25.1); use (system-name) instead

I've used (system-name) since at least 2011, so it must have been
around quite a while.

3 years agolib/message.cc: use view number to invalidate cached metadata
David Bremner [Fri, 24 Feb 2017 18:33:07 +0000 (14:33 -0400)]
lib/message.cc: use view number to invalidate cached metadata

Currently the view number is incremented by notmuch_database_reopen

3 years agolib: handle DatabaseModifiedError in _n_message_ensure_metadata
David Bremner [Sat, 18 Feb 2017 01:28:05 +0000 (21:28 -0400)]
lib: handle DatabaseModifiedError in _n_message_ensure_metadata

The retries are hardcoded to a small number, and error handling aborts
than propagating errors from notmuch_database_reopen. These are both
somewhat justified by the assumption that most things that can go
wrong in Xapian::Database::reopen are rare and fatal. Here's the brief
discussion with Xapian upstream:

   24-02-2017 08:12:57 < bremner> any intuition about how likely
      Xapian::Database::reopen is to fail? I'm catching a
      DatabaseModifiedError somewhere where handling any further errors is
      tricky, and wondering about treating a failed reopen as as "the
      impossible happened, stopping"

   24-02-2017 16:22:34 < olly> bremner: there should not be much scope for
    failure - stuff like out of memory or disk errors, which are probably a
    good enough excuse to stop

3 years agolib: add _notmuch_database_reopen
David Bremner [Fri, 24 Feb 2017 01:38:24 +0000 (21:38 -0400)]
lib: add _notmuch_database_reopen

The main expected use is to recover from a Xapian::DatabaseChanged
exception.

3 years agotest: add known broken test for uncaught DatabaseModifiedError
David Bremner [Sat, 25 Feb 2017 02:57:40 +0000 (22:57 -0400)]
test: add known broken test for uncaught DatabaseModifiedError

There are several of these to track down, but one that is in quite a
few code paths is _notmuch_message_ensure_metadata.

3 years agodebian: note read-after-free fix in d/changelog
David Bremner [Sat, 25 Feb 2017 12:48:10 +0000 (08:48 -0400)]
debian: note read-after-free fix in d/changelog

3 years agoNEWS: document read-after-free bugfix
David Bremner [Sat, 25 Feb 2017 12:45:43 +0000 (08:45 -0400)]
NEWS: document read-after-free bugfix

3 years agoMerge branch 'release'
David Bremner [Thu, 23 Feb 2017 13:08:15 +0000 (09:08 -0400)]
Merge branch 'release'

Merge in g_hash_table read-after-free fix

3 years agoversion: bump to 0.23.6
David Bremner [Thu, 23 Feb 2017 13:03:10 +0000 (09:03 -0400)]
version: bump to 0.23.6

python and debian bumped at the same time

3 years agotest: replaced use of python with $NOTMUCH_PYTHON (twice)
Tomi Ollila [Wed, 4 Jan 2017 21:33:15 +0000 (23:33 +0200)]
test: replaced use of python with $NOTMUCH_PYTHON (twice)

$NOTMUCH_PYTHON is sourced from sh.config, configured by
./configure and stated to be used as:

"Name of python command to use in configure and the test suite."

3 years agoemacs: company: remove two build warnings
Mark Walters [Sat, 28 Jan 2017 10:19:58 +0000 (10:19 +0000)]
emacs: company: remove two build warnings

This adds two defvars to avoid two build warnings in the notmuch emacs
code. These were both introduced (by me) in commit 827c28a0.

3 years agoemacs: fully add the notmuch-address customize group
Mark Walters [Sat, 28 Jan 2017 10:38:27 +0000 (10:38 +0000)]
emacs: fully add the notmuch-address customize group

We now have several customizable options for address completion. There
is a customize group notmuch-address but it only contains one of these
options. Add all the others, and make it part of the notmuch customize
group.

3 years agolib: make _notmuch_message_ensure_property_map static
David Bremner [Sat, 18 Feb 2017 14:45:48 +0000 (10:45 -0400)]
lib: make _notmuch_message_ensure_property_map static

It's not called outside message.cc

3 years agolib: make _notmuch_message_ensure_metadata static
David Bremner [Sat, 18 Feb 2017 14:45:44 +0000 (10:45 -0400)]
lib: make _notmuch_message_ensure_metadata static

It's not called anywhere outside message.cc.

3 years agolib: fix g_hash_table related read-after-free bug
David Bremner [Wed, 22 Feb 2017 10:17:47 +0000 (06:17 -0400)]
lib: fix g_hash_table related read-after-free bug

The two g_hash_table functions (insert, add) have different behaviour
with respect to existing keys. g_hash_table_insert frees the new key,
while g_hash_table_add (which is really g_hash_table_replace in
disguise) frees the existing key. With this change 'ref' is live until
the end of the function (assuming single-threaded access to
'hash'). We can't guarantee it will continue to be live in the
future (i.e. there may be a future key duplication) so we copy it with
the allocation context passed to parse_references (in practice this is
the notmuch_message_t object whose parents we are finding).

Thanks to Tomi for the simpler approach to the problem based on
reading the fine glib manual.

3 years agoMerge branch 'release'
David Bremner [Tue, 21 Feb 2017 12:16:07 +0000 (08:16 -0400)]
Merge branch 'release'

Merge changes to use gpgconf --create-socketdir

3 years agodebian: start changelog for 0.23.6
David Bremner [Tue, 21 Feb 2017 11:51:44 +0000 (07:51 -0400)]
debian: start changelog for 0.23.6

3 years agoNEWS: news for gpgconf use in test suite
David Bremner [Tue, 21 Feb 2017 11:50:05 +0000 (07:50 -0400)]
NEWS: news for gpgconf use in test suite

3 years agotest: use gpgconf --create-socketdir if available
David Bremner [Tue, 14 Feb 2017 21:42:39 +0000 (16:42 -0500)]
test: use gpgconf --create-socketdir if available

This enables the shortened socket pathes in /run or equivalent. The
explicit call to gpgconf is needed for nonstandard GNUPGHOME settings.

(amended according to id:m2fujatr4k.fsf@guru.guru-group.fi)

3 years agoconfigure: add test for gpgconf --create-socketdir
David Bremner [Tue, 14 Feb 2017 21:42:38 +0000 (16:42 -0500)]
configure: add test for gpgconf --create-socketdir

This is primarily intended for use in the test suite (since notmuch
builds fine without gnupg installed). Thus we only write the variable
to sh.config.

3 years agolib: Let Xapian manage the memory for FieldProcessors
David Bremner [Fri, 17 Feb 2017 03:07:50 +0000 (23:07 -0400)]
lib: Let Xapian manage the memory for FieldProcessors

It turns out this is exactly what release() is for; Xapian will
deallocate the objects when it's done with them.

3 years agolib: merge internal prefix tables
David Bremner [Fri, 17 Feb 2017 03:07:49 +0000 (23:07 -0400)]
lib: merge internal prefix tables

Replace multiple tables with some flags in a single table. This makes
the code in notmuch_database_open_verbose a bit shorter, and it should
also make it easier to add other options to fields, e.g. regexp
searching.

3 years agoemacs: address: save hash
Mark Walters [Sun, 20 Nov 2016 16:50:34 +0000 (16:50 +0000)]
emacs: address: save hash

This allows the user to save the address hash so that it is much
faster for the first completion after a restart. This defaults to off
as there are privacy implications to saving this information.

The code tries hard to avoid overwriting the wrong file. It also notes
if changes have been made to any of the relevant user settings, so
that the user does not get surprising results (i.e., outdated options
being used). Finally it stores some version information so that is
easy for us to update the format of the save file.

3 years agoemacs: address: move address-full-harvest-finished to a function
Mark Walters [Sun, 20 Nov 2016 16:50:33 +0000 (16:50 +0000)]
emacs: address: move address-full-harvest-finished to a function

This makes the code access notmuch-address-full-harvest-finished via a
helper function, notmuch-address--harvest-ready. Later we will use
this to check whether we can load the harvest instead of regenerating
it.

3 years agocli: don't call _entry_in_ignore_list twice in count files debug
Jani Nikula [Sat, 17 Dec 2016 16:28:43 +0000 (18:28 +0200)]
cli: don't call _entry_in_ignore_list twice in count files debug

Split file ignores in count_files to fixed and user configured in
order to not have to call _entry_in_ignore_list twice when debugging
is enabled. Minor detail.

3 years agotest: wrap 'wc -l' results in arithmetic evaluation to strip whitespace
Tomi Ollila [Wed, 4 Jan 2017 22:35:23 +0000 (00:35 +0200)]
test: wrap 'wc -l' results in arithmetic evaluation to strip whitespace

Some new unwrapped 'wc -l's have been added since Jani's 60e79e3a9f1c8

3 years agolib: optimize counting documents
David Bremner [Thu, 12 Jan 2017 02:01:15 +0000 (22:01 -0400)]
lib: optimize counting documents

From #xapian

olly> bremner: btw, i noticed notmuch count see ms to request all the documents and then ignores them

bremner> hmm. There's something funny about the way that notmuch uses matches in general iirc

olly> it should be able to do: mset = enquire.get_mset (0, 0, notmuch->xapian_db->get_doccount ());
...
olly> get_matches_estimated() will be exact because check_at_least is the size of the database

3 years agodocs: fix notmuch_message_properties_value documentation
Steven Allen [Fri, 13 Jan 2017 17:51:02 +0000 (09:51 -0800)]
docs: fix notmuch_message_properties_value documentation

It returns the value, not the key.

3 years agoMerge tag '0.23.5'
David Bremner [Mon, 9 Jan 2017 10:47:30 +0000 (06:47 -0400)]
Merge tag '0.23.5'

notmuch 0.23.5 release

3 years agoconfigure: remove leftover byte order test cleanup
Jani Nikula [Sun, 8 Jan 2017 17:09:45 +0000 (19:09 +0200)]
configure: remove leftover byte order test cleanup

Removing the removal of byteorder configure test files was overlooked
in commit 5a957c3f33e6 ("build & util: drop byte order configure check
and endian util header"). Finish the job.

3 years agoNEWS,debian: set date for release 0.23.5 debian/0.23.5-1
David Bremner [Mon, 9 Jan 2017 10:25:01 +0000 (06:25 -0400)]
NEWS,debian: set date for release

3 years agobuild & util: drop byte order configure check and endian util header
Jani Nikula [Fri, 6 Jan 2017 21:29:28 +0000 (23:29 +0200)]
build & util: drop byte order configure check and endian util header

With the removal of the embedded libsha1, we lost the first and last
user of the platform byte order checks. Remove them from configure,
and remove the endian util header.

3 years agolib: use glib for sha1 digests instead of embedding libsha1
Jani Nikula [Fri, 6 Jan 2017 21:29:27 +0000 (23:29 +0200)]
lib: use glib for sha1 digests instead of embedding libsha1

We already depend on glib both directly and indirectly (via gmime). We
might as well make use of its facilities. Drop the embedded libsha1
and use glib for sha1 digests.

3 years agotest: allow user to choose which gdb to run tests with
Tomi Ollila [Sat, 7 Jan 2017 09:47:27 +0000 (11:47 +0200)]
test: allow user to choose which gdb to run tests with

The variable used for selecting gdb is TEST_GDB, consistent with
TEST_CC and TEST_EMACS{,CLIENT}.

3 years agoMerge branch 'release'
David Bremner [Sun, 8 Jan 2017 12:42:26 +0000 (08:42 -0400)]
Merge branch 'release'

These are the (tentative) changes for 0.23.5

3 years agodocs: add 2017 to copyright years
David Bremner [Sun, 8 Jan 2017 12:35:17 +0000 (08:35 -0400)]
docs: add 2017 to copyright years

3 years agobump version to 0.23.5
David Bremner [Sun, 8 Jan 2017 12:32:56 +0000 (08:32 -0400)]
bump version to 0.23.5

3 years agodebian: changelog for 0.23.5-1
David Bremner [Sun, 8 Jan 2017 12:31:28 +0000 (08:31 -0400)]
debian: changelog for 0.23.5-1

3 years agoNEWS for 0.23.5
David Bremner [Sun, 8 Jan 2017 12:30:08 +0000 (08:30 -0400)]
NEWS for 0.23.5

3 years agolib: fix the todo comment placement on NOTMUCH_STATUS_XAPIAN_EXCEPTION
Jani Nikula [Sat, 17 Dec 2016 11:05:40 +0000 (13:05 +0200)]
lib: fix the todo comment placement on NOTMUCH_STATUS_XAPIAN_EXCEPTION

The todo comment got separated from the status it's related to at
commit 3f32fd8a1c06 ("Add missing comment for
NOTMUCH_STATUS_READONLY_DATABASE."). Later, commit b65ca8e0ba5e ("lib:
modify notmuch.h for automatic document generation") moved it, but to
the wrong place. Fix the location.

3 years agocli/config: rename name to key in _config_set_list
Jani Nikula [Sat, 17 Dec 2016 16:20:27 +0000 (18:20 +0200)]
cli/config: rename name to key in _config_set_list

The other getters and setters speak of key, follow suit. No functional
changes.

3 years agocli/config: allocate config using talloc_zero
Jani Nikula [Sat, 17 Dec 2016 16:20:26 +0000 (18:20 +0200)]
cli/config: allocate config using talloc_zero

Do not initialize each field separately. It's more robust to allocate
the config with zero initialization, and only set the non-zero
defaults individually.

3 years agoconfigure: fix $prefix expansion for libdir_expanded
Tomi Ollila [Sun, 1 Jan 2017 17:20:30 +0000 (19:20 +0200)]
configure: fix $prefix expansion for libdir_expanded

Since the sed expansion line which did $prefix expansion for
libdir_expanded was changed from the legacy `...` format to the
new $(...) expression, the subtle backslash expansion change went
unnoticed -- \\$ which used to escape '$' now escapes '\' and the
following '$prefix' was attempted to expand as a variable. So
changing \\$ to \$ fixes this.

Also, replaced echo with printf %s -- echo does expansions of its own.

While at it, the following 2 inconsistencies were fixed:
 1) the /g flag was removed from first expression; second didn't have it
 2) first expression did not end with /, so "dropped" it from second

 configure | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

3 years agoMerge branch 'release'
David Bremner [Wed, 28 Dec 2016 23:41:20 +0000 (08:41 +0900)]
Merge branch 'release'

merge 0.23.4 release

3 years agoversion: update to 0.23.4 0.23.4 debian/0.23.4-1
David Bremner [Sat, 24 Dec 2016 08:52:11 +0000 (17:52 +0900)]
version: update to 0.23.4

3 years agoset date in NEWS and debian/changelog
David Bremner [Sat, 24 Dec 2016 08:48:16 +0000 (17:48 +0900)]
set date in NEWS and debian/changelog

pentultimate prepartions for release

3 years agodebian: remove setting of emacslispdir
David Bremner [Thu, 22 Dec 2016 23:34:30 +0000 (08:34 +0900)]
debian: remove setting of emacslispdir

With dh-elpa, this setting is no longer used

3 years agodebian: changelog stanza for 0.23.4-1
David Bremner [Sun, 18 Dec 2016 11:48:48 +0000 (20:48 +0900)]
debian: changelog stanza for 0.23.4-1

3 years agonotmuch-config: replace config reading function
Ioan-Adrian Ratiu [Sat, 10 Dec 2016 19:54:07 +0000 (21:54 +0200)]
notmuch-config: replace config reading function

Config files are currently read using glib's g_key_file_load_from_file
function which is very inconvenient because it's limited by design to read
only from "regular data files" in a filesystem. Because of this limitation
notmuch can't read configs from pipes, fifos, sockets, stdin, etc. Not even
"notmuch --config=/dev/stdin" works:

Error reading configuration file /dev/stdin: Not a regular file

So replace g_key_file_load_from_file with g_key_file_load_from_data which
gives us much more freedom to read configs from multiple sources.

This also helps the more security sensitive users: If someone has private
information in the config file, it can be encrypted on disk, then decrypted
in RAM and passed through a pipe directly to notmuch without the use of
intermediate plain text files.

Signed-off-by: Ioan-Adrian Ratiu <adi@adirat.com>
3 years agocli: abstract config file reading to a separate function
Jani Nikula [Sat, 10 Dec 2016 19:54:06 +0000 (21:54 +0200)]
cli: abstract config file reading to a separate function

Simplify and fix the coding style while at it.

3 years agoMerge branch 'release'
David Bremner [Thu, 15 Dec 2016 12:57:10 +0000 (21:57 +0900)]
Merge branch 'release'

Add back in no-display parameter by hand.

3 years agoNEWS: announce autoload fix
David Bremner [Thu, 8 Dec 2016 00:45:52 +0000 (20:45 -0400)]
NEWS: announce autoload fix

3 years agoemacs: restore autoload cookie for notmuch-search
David Bremner [Thu, 8 Dec 2016 00:45:51 +0000 (20:45 -0400)]
emacs: restore autoload cookie for notmuch-search

The cookie only applies to next form, so in the previous location it
applied to the put, rather than the function.

3 years agoNEWS: news for notmuch-insert error handling
David Bremner [Wed, 7 Dec 2016 11:27:13 +0000 (07:27 -0400)]
NEWS: news for notmuch-insert error handling

3 years agocli/insert: document the use of EX_TEMPFAIL
David Bremner [Wed, 7 Dec 2016 11:27:12 +0000 (07:27 -0400)]
cli/insert: document the use of EX_TEMPFAIL

3 years agoMerge branch 'release'
David Bremner [Wed, 7 Dec 2016 11:28:46 +0000 (07:28 -0400)]
Merge branch 'release'

Merge changes for notmuch-insert error handling, debian bugfix release

3 years agocli/insert: return EX_TEMPFAIL for some errors
David Bremner [Mon, 28 Nov 2016 03:01:42 +0000 (23:01 -0400)]
cli/insert: return EX_TEMPFAIL for some errors

Attempt to distinguish between errors indicating misconfiguration or
programmer error, which we consider "permanent", in the sense that
automatic retries are unlikely to be useful, and those indicating
transient error conditions. We consider XAPIAN_EXCEPTION transient
because it covers the important special case of locking failure.

3 years agocli/insert: delay database open until after writing mail file
David Bremner [Sun, 27 Nov 2016 15:24:58 +0000 (11:24 -0400)]
cli/insert: delay database open until after writing mail file

The idea is to get the mail written to disk, even if we can't open the
database (e.g. because some other process has a write lock, and notmuch
is compiled for non-blocking opens).

3 years agotest: gdb insert: redirect input inside gdb script
Tomi Ollila [Mon, 28 Nov 2016 21:07:08 +0000 (23:07 +0200)]
test: gdb insert: redirect input inside gdb script

Running `gdb command < input` is not as reliable way to give input
to the command (some installations of gdb consume it). Use "set args"
gdb command to have input redirected at gdb 'run' time.

3 years agodebian: changelog stanza for 0.23.3-3
David Bremner [Mon, 5 Dec 2016 12:26:28 +0000 (08:26 -0400)]
debian: changelog stanza for 0.23.3-3

3 years agodebian: disable gdb using tests on kfreebsd-*
David Bremner [Mon, 5 Dec 2016 12:22:57 +0000 (08:22 -0400)]
debian: disable gdb using tests on kfreebsd-*

gdb seems broken there.

3 years agotest: check the handling of files vanishing between scandir and indexing
Jani Nikula [Sun, 27 Nov 2016 09:59:34 +0000 (11:59 +0200)]
test: check the handling of files vanishing between scandir and indexing

Add a file for scandir to find, but use gdb to remove it before it
gets indexed.

3 years agotest: clean up corpora.email
David Bremner [Sun, 27 Nov 2016 12:04:38 +0000 (08:04 -0400)]
test: clean up corpora.email

When we renamed corpus to corpora, we missed a spot.

3 years agoMerge branch 'release'
David Bremner [Thu, 1 Dec 2016 12:57:48 +0000 (08:57 -0400)]
Merge branch 'release'

3 years agodebian: fix transition package for notmuch-emacs
David Bremner [Thu, 1 Dec 2016 12:08:14 +0000 (08:08 -0400)]
debian: fix transition package for notmuch-emacs

The current version is more a transition to nowhere.

3 years agobuild: update the desktop database after installing the desktop file
Jani Nikula [Sat, 26 Nov 2016 09:43:49 +0000 (11:43 +0200)]
build: update the desktop database after installing the desktop file

This makes the option to choose Notmuch as mailto: handler show up in
the desktop environment settings. Ignore errors.

3 years agobuild: install notmuch-emacs-mua.desktop file with emacs
Jani Nikula [Sat, 26 Nov 2016 09:37:17 +0000 (11:37 +0200)]
build: install notmuch-emacs-mua.desktop file with emacs

There is really no need to have a separate install target for the
desktop file. Just install the desktop file with emacs, with a
configure option to opt out.

3 years agodoc: whitespace cleanup
Tomi Ollila [Sat, 26 Nov 2016 22:25:57 +0000 (00:25 +0200)]
doc: whitespace cleanup

In most part, our .rst documents are indented with 8 spaces instead
of tabs. Bring the rest of the lines to the same format.
Also, on one (supposedly empty) line, trailing spaces were removed.

3 years agoMerge tag '0.23.3'
David Bremner [Sun, 27 Nov 2016 12:09:17 +0000 (08:09 -0400)]
Merge tag '0.23.3'

notmuch 0.23.3 release

3 years agoNEWS: fix date and title 0.23.3
David Bremner [Sun, 27 Nov 2016 01:54:10 +0000 (21:54 -0400)]
NEWS: fix date and title

3 years agoMerge branch 'release'
David Bremner [Sat, 26 Nov 2016 12:43:31 +0000 (08:43 -0400)]
Merge branch 'release'

Add Jani's fix for notmuch-new

3 years agoNEWS: note disappearing files fix for notmuch new
David Bremner [Sat, 26 Nov 2016 12:41:47 +0000 (08:41 -0400)]
NEWS: note disappearing files fix for notmuch new

3 years agoNEWS: set date
David Bremner [Sat, 26 Nov 2016 12:39:32 +0000 (08:39 -0400)]
NEWS: set date