| Age | Commit message (Collapse) | Author |
|---|
|
As can be seen from the included tests, this currently only works for
fields without other field processors (in particular regex supporting
fields seem to break it).
|
|
Xapian 2.0 supports extended wildcard queries in the query parser. To
quote the NEWS file:
WILDCARD_PATTERN_MULTI (which supports `*` wildcards anywhere in
the term, not just at the end as previously),
WILDCARD_PATTERN_SINGLE (which supports `?` wildcards matching a
single character), and WILDCARD_PATTERN_GLOB which enables both of
these.
This commit does not add any functionality, just detects if this
support is there in Xapian.
|
|
In [1], Anton mentioned encountering a race condition in the build
system, and suggested this fix.
[1]: id:177149756401.20239.4570498044656906075@lain.khirnov.net
|
|
This allows tests to pass on systems where the readelf executable is
prefixed with a target triple.
Signed-off-by: Lucio Sauer <watermanpaint@posteo.net>
|
|
Use `when-let*' instead of `when-let'. Use quoted
`font-lock-warning-face' instead of obsolete variable.
|
|
This innocuous looking omission actually breaks the Debian packaging
due to 'dh_missing'.
|
|
In [1], Michael pointed out we should s/proved/provided/.
[1]: id:CAA19uiQvyg_fF4z+92D6Zhn_DOx+VC8thct7MioxCsEZ78CYVg@mail.gmail.com
|
|
While it is not intended to be invoked directly, the user still needs
to know how to use it via git commands.
|
|
|
|
In cases where a given 'export' both adds and deletes messages, we may
not know what messages are actually missing until the end of
processing. We thus do single pass of all remaining message-ids, and
report any that are still missing at the end of the "export"
operation.
|
|
There are two main possibilities. One is explicit delete ('D' command
in the stream from git fast-export) and one is files disappearing
between commits. It is less clear the latter can happen in well formed
sequence of commits, but it could result e.g. from manual changes to
the repo.
|
|
Two (sub)features are stubbed out in this initial implementation:
deleting messages (as opposed to tags), and missing messages. There
are two corresponding tests marked as broken in T860-git-remote.sh.
A third test passes with the stub, which is maybe not ideal, but at
least it acts as a regression test.
|
|
The output in default.import is based on a modified version
of Felipe's git-remote-nm with Blake2 hashing replaced by SHA1
(for portability). This enable fetch/pull/clone, so test that as well.
|
|
This is closely based on git-remote-nm (in ruby) by Felipe Contreras.
Initially just implement the commands 'capabilities' and 'list'. This
isn't enough to do anything useful but we can run some simple unit tests. Testing
of URL passing will be done after clone (import command) support is
added.
|
|
Moving these functions to libnotmuch_util will allow re-use from
either multiple CLI compilation units or from the library. To avoid
future surprises, replace printing to stderr with the usual status
string mechanism.
|
|
The precise function will be explained in a future update to
notmuch-config(1).
|
|
This quirk is somewhat unexpected, since flags won't be updated as
expected, but subsequence calls to notmuch-new(1) won't update the
index, giving an impression of it being stale.
|
|
pkg-config could be prefixed in some env.
Signed-off-by: Nicolas PARLANT <nicolas.parlant@parhuet.fr>
|
|
|
|
notmuch 0.40 release
|
|
|
|
|
|
Avoid spurious underlining in markdown output.
|
|
|
|
Basically transcribe a bit more of the corresponding commit message.
|
|
This seems to have been copied accidentally from another part of the
file.
|
|
`message-cite-reply-position' affects the position of the email body
and signature relative to the email citation in email replies.
Test each of the three possible values of
`message-cite-reply-position'. If `message-cite-reply-position' is
'traditional or 'below, place the email signature below the
citation. If `message-cite-reply-position' is 'above, place the email
signature above the citation (as in Gmail-style email replies).
|
|
Previously, when composing email replies with `notmuch-mua-new-mail',
the email signature would always be placed below the email citation.
However, there are two message.el user options that affect the
position of the signature and email body:
`message-cite-reply-position` and `message-cite-style-gmail'.
Previously, neither of these user options were respected.
Respect these user options. If `message-cite-reply-position' is
'traditional or 'below, place the email signature below the citation
(the previous behavior is retained). If `message-cite-reply-position'
is 'above, place the email signature above the citation (like in
Gmail-style email replies).
`message-cite-style-gmail' may specify a value for
`message-cite-reply-position'. If it does, that value takes precedence
over `message-cite-reply-position'.
|
|
`notmuch-tree-filter-by-tag' and `notmuch-search-filter-by-tag' are
described, in their docstrings, as matching all the messages in the
current view that are ALSO tagged by a given tag. However, since the
current view's search query was not logically grouped together (with
parentheses), the actual behavior was different for sufficiently
complex search queries.
Fix this by grouping the current view's search query in parentheses
first.
|
|
[dgit (14.4) quilt-fixup]
|
|
|
|
|
|
Commit 3af10307d started a subsection on message level actions, but
the author (yours truly) mistook their customized behaviour for the
default.
|
|
|
|
Workaround for apparent toolchain regression on armhf.
|
|
[dgit (14.2) quilt-fixup]
|
|
|
|
This reflects the addition of a new API functions,
notmuch_{messages,threads}_status
|
|
Since the arch no longer exists, it is no longer helpful to exclude it.
|
|
|
|
|
|
|
|
|
|
As reported by "lh" on IRC.
|
|
In the search box (and recent searches), notmuch-search-hide-excluded
was not being passed to notmuch-search. This means that excluded
messages were always shown in the results. Fix this by passing the
extra argument.
|
|
One of the places notmuch-search is called did not get the new
'hide-excluded' argument.
|
|
'hide-excluded' argument.
|
|
It is wrong most of the time in _notmuch_message_remove_terms, but
that function is too low level to know how to call
_n_m_invalidate_metadata with the right argument, at least not without
more extensive changes. This change merely makes the current behaviour
more obvious, since the other calls cannot have relied on metadata
being invalidated.
|
|
Previously when installed with 'make install', autoloads in sources
were not scraped. So users still had to load the full library
beforehand in order to use Notmuch.
Now we generate and install such a file for 'make install-emacs'. With
this change, users can replace in their Emacs configuration:
(require 'notmuch)
with a much cheaper:
(require 'notmuch-autoloads)
but are still able to use 'M-x notmuch'.
Note that this new file is not included in the 'make elpa' package, so
it does not break for those users, who have already been enjoying
autoloads generated by 'package.el'.
Besides, this should not break packaging for most distros. Even if
they have generated and maintained such a file using their package
managers, Emacs does not choke on an existing output file when
scraping autoloads.
Re: thread starting at id:87o75yl4u5.fsf@pengjiz.com
Amended-By: db; add comment about autoload-excludes
|
|
Throughout the project's files, contributors have implicitly been
ending their sentences with a period and a single space. Set
`sentence-end-double-space` accordingly.
|
