Since the error field is unused by the emacs front end, no changes are
needed other than bumping the format version number.
As it is, this is a bit overengineered, but it will reduce duplication
when we support gmime 3.0
This is another case where the behaviour of gmime-2.6 and gmime-3.0
seems to differ. It may be that we prefer the more lax parsing of the
previous version, but that should be tested separately.
Since gnupg 2.1.20, gpg-agent no longer shut itself down when
$GNUPGHOME directory is removed.
Add exit hooks to the test modules which execute `gpgconf --kill all`
Add exit hooks to execute `gpgconf --kill all` in the modules that
create $GNUPGHOME for gpg to work with.
New function at_exit_function registers given function to be called
at script termination.
Functions so registered are called in the reverse order of their
registration; no arguments are passed.
Function is called only once; re-adding with function name already
registered will remove previous registration.
New function rm_exit_function can be used to remove registration.
Modules (and possibly test-lib.sh functions) in future commits will
register such functions.
The dynamic generation of the linker version script for libnotmuch
exports has grown rather complicated.
Reverse the visibility control by hiding symbols by default using
-fvisibility=hidden, and explicitly exporting symbols in notmuch.h
using #pragma GCC visibility. (We could also use __attribute__
((visibility ("default"))) for each exported function, but the pragma
is more convenient.)
The above is not quite enough alone, as it would "leak" a number of
weak symbols from Xapian and C++ standard library. Combine it with a
small static version script that filters out everything except the
notmuch_* symbols that we explicitly exposed, and the C++ RTTI
typeinfo symbols for exception handling.
Finally, as the symbol hiding test can no longer look at the generated
symbol table, switch the test to parse the functions from notmuch.h.
From a UI perspective this looks similar to what was already provided
for from, subject, and mid, but the implementation is quite
different. It uses the database's list of terms to construct a term
based query equivalent to the passed regular expression.
'quite' on IRC reported that notmuch new was grinding to a halt during
initial indexing, and we eventually narrowed the problem down to some
html parts with large embedded images. These cause the number of terms
added to the Xapian database to explode (the first 400 messages
generated 4.6M unique terms), and of course the resulting terms are
not much use for searching.
The second test is sanity check for any "improved" indexing of HTML.
These 210 messages are in several long threads, which is good for
testing our threading code, and may be useful just as a larger test
corpus in the future.
The non-field processor behaviour is is convert the corresponding
queries into a search for the unprefixed terms. This yields pretty
surprising results so I decided to generate a query that would match
the terms (i.e. none with that prefix) generated for an empty header.
Make test-lib-common.sh load test-lib-<$PLATFORM>.sh to create
additional shim for platform specifics.
Use test-lib-FREEBSD.sh to call GNU utilities instead of native ones.
- amended by db following Tomi's suggestions
The argument is that if the string passed to the field processor has
no spaces, then the added quotes won't have any benefit except for
disabling wildcards. But disabling wildcards doesn't seem very useful
in the normal Xapian query parser, since they're stripped before
generating terms anyway. It does mean that the query 'from:"foo*"' will
not be precisely equivalent to 'from:foo' as it is for the non
field-processor version.
This was broken by the addition of regexp searching. The detection of
wildcards is not currently done in the recursive call to parse_query,
because of quoting issues.
This function was deprecated in notmuch 0.21. We re-use the name for
a status returning version, and deprecate the _st name. One or two
remaining uses of the (removed) non-status returning version fixed at
the same time
Apparently some systems (MacOS?) have a system library called libutil
and the name conflict causes problems. Since this library is quite
notmuch specific, rename it to something less generic.
Often Linux systems are shipped with wget(1) by default (and no curl).
Many BSDs, macOS, and e.g. some Linux minimal/container images
comes with curl(1) (and no wget).
Attempting to download with curl if wget is not available increases
the likelihood for this to succeed.
For reasons not completely understood at this time, gmime (as of
2.6.22) is returning a date before 1900 on bad date input. Since this
confuses some other software, we clamp such dates to 0,
i.e. 1970-01-01.
Some gdb python exceptions on some os environments (e.g. macOS
Sierra, non-codesigned gdb) do not make gdb exit (but to drop down
to nonexistent command line?).
Mitigate this chance by explict SystemExit on all exceptions.
The contents of output file 'gdb.out' is unchanged.
Unify the subtests by requiring test_begin_subtest before
test_expect_code. (Similar change for test_expect_success has already
been done.)
This increases clarity in the test scripts by having a separate line
for the start of the subtest with the heading, and makes it possible
to simplify the test infrastructure by making all subtests similar.
Unify the subtests by requiring test_begin_subtest before
test_expect_success. (Similar change for test_expect_code will
follow.)
This increases clarity in the test scripts by having a separate line
for the start of the subtest with the heading, and makes it possible
to simplify the test infrastructure by making all subtests similar.
The only place where we use the implicit prereq check is T000-basic.sh
where we check that it works. It's an added complication that we don't
use. Remove it.
The test_have_prereq function can still be used for the same effect in
subtests that use test_begin_subtest. For now, this will make it
impossible to have prereqs in one-line subtests that don't require
test_begin_subtest. This will be fixed in follow-up work.
Known broken tests are, well, known broken. Do not print the result
diff for them unless V=1 is specified. Now that the test description
is printed also when known broken tests fail, the user can also skip
to running the individual failing tests.
Headers of more than 998 characters should be folded when sending.
However, until recently, emacs did not do this.
This adds a (known broken) test for this when sending messages in
emacs. We will backport the fix to notmuch-emacs in the next
changeset.
We filter added exclude at add time, rather than modifying the query by
count search. As noted in the comments, there are several ignored
conditions here.
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.
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).
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.
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.
This reverts commit e7b88e8b0a.
It turns out that this does not work well in environments without a
running systemd (or some other provider of /run/user)
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.
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.
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
$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."
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)
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.
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.
GnuPG 2.1.16 is now injecting the full issuer fingerprint in its
signatures, which makes them about 32 octets larger when
ascii-armored.
This change in size means that the size of the MIME parts will vary
depending on the version of gpg that the user has installed. at any
rate, the signature part should be non-zero (this is true for
basically any MIME part), so we just test for that instead of an exact
size.
emacs24 and emacs23 have different secure tag defaults: in particular,
mml-secure-message-sign only signs the part on emacs23 but the whole
message on emacs24. This difference makes one of the draft tests fail
(which causes a cascade of later failures) on emacs23. It seems that
travis uses emacs23 so it is useful to fix this.
We do this by forcing the whole message to be signed in either case --
the code snippet is extracted from mml-secure-message-sign on emacs24.
Provide functionality to resume editing a message previously saved with
notmuch-draft-save, including decoding the X-Notmuch-Emacs-Secure
header.
Resume gets the raw file from notmuch and using the emacs function
mime-to-mml reconstructs the message (including attachments).
'e' is bound to resume a draft from show or tree mode.
This provides initial support for postponing in the emacs frontend;
resuming will follow in a later commit. On saving/postponing it uses
notmuch insert to put the message in the notmuch database
Current bindings are C-x C-s to save a draft, C-c C-p to postpone a
draft (save and exit compose buffer).
Previous drafts get tagged deleted on subsequent saves, or on the
message being sent.
Each draft gets its own message-id, and we use the namespace
draft-.... for draft message ids (so, at least for most people, drafts
are easily distinguisable).
Moved the 2 basename(1) executions to the test failure branch in
test_expect_equal_file ().
The output of basename(1) executions in function test_expect_equal_file ()
are only used when tests fails -- when all tests pass these 2 basename(1)
executions are no longer done at all.
Thanks to Lucas (id:147263183913.27784.12274024193186585889@mbp) for the
bug report and the test case.
I decided to use the python version because the python bindings could
use more exercise.
In case of the test script is to be relaunced under valgrind, or --tee
is requested, use the $BASH shell variable to locate the command
interpreter. The $SHELL variable is re-set by non-interactive shells
so in case the shell uses some other shell (e.g. zsh) for interactive
use these bash scripts continue to work.
This is a strange corner case where the removing of the user's address
from the To: header does the wrong thing. If we think it is
worth (eventually) fixing, this test can serve as a reminder.
We want to be able to query the properties directly, like:
notmuch count property:foo=bar
which should return a count of messages where the property with key
"foo" has value equal to "bar".
This stops the (usually incorrect) sigstatus and encstatus buttons
appearing when replying in emacs, and updates the test suite to match.
Overriding the status button functions is a little unusual but much
less intrusive than passing an argument all the way down the call
chain. It also makes it clear exactly what it does.
We also hide the application/pgp-encrypted part as it can only contain
"Version: 1". We do this in notmuch show, which means it also happens
when replying.
Pass in GMimeMessage to simplify To/Cc/Bcc headers. We'll eventually
remove the notmuch message passing altogether, but keep both for now
to not make too big changes at once.
Getting the headers from GMimeMessage using GMime functions fixes the
error on duplicate Cc headers reported by Daniel Kahn Gillmor
<dkg@fifthhorseman.net> in id:87d1ngv95p.fsf@alice.fifthhorseman.net.
Get rid of an intermediate function.
The small annoyance is the ownership differences in the address lists.
As Daniel Kahn Gillmor <dkg@fifthhorseman.net> reports in
id:87d1ngv95p.fsf@alice.fifthhorseman.net, notmuch show combines
multiple Cc: fields into one, while notmuch reply does not. While such
messages are in violation of RFC 5322, it would be reasonable to
expect notmuch to be consistent. Add a known broken test to document
this expectation.
This also starts a new "broken" corpus for messages which are broken.
Details:
The original message is formatted using the message printing in
notmuch-show.c. For Cc:, it uses g_mime_message_get_recipients(),
which apparently combines all Cc: fields into one internally.
The addresses in the reply headers, OTOH, are based on headers queried
through libnotmuch. It boils down to g_mime_object_get_header() in
lib/message-file.c, which returns only the first occurence of header.
We largely use the corpus under test/corpus for
testing. Unfortunately, many of our tests have grown to depend on
having exactly this set of messages, making it hard to add new message
files for testing specific cases.
We do use a lot of add_message from within the tests, but it's not
possible to use that for adding broken messages, and adding several
messages at once can get unwieldy.
Move the basic corpus under tests/corpora/default, and make it
possible to add new, independent corpora along its side. This means
tons of renames with a few tweaks to add_email_corpus function in
test-lib.sh to let tests specify which corpus to use.
Originally the intent was to make the test more robust against changing
test keys. It turns out that (unscientifically) gpg --with-colons output
changes more often than our test key. Rather than making the script more
complex, just hard code the fingerprint.
This fixes Debian bug #847013; I expect similar test failures as other
distros adopt gnupg 2.1.15
With all the preparation it is now simple to add the actual insert
code. Since insert can fail for many reasons we let the user decide
interactively deal with it.
We modify test-lib.el to set file fcc, so that all the old tests and
emacs_fcc_message from test-lib.sh still work
gdb sometimes writes warnings to stdout, which we don't need/want, and
for some reason --batch-silent isn't enough to hide. So in this commit
we write them to a log file, which is probably better for debugging
anyway. To see an illustrative test failure before this change, run
% make
% touch notmuch-count.c
% cd test && ./T060-count.sh
(cherry picked from commit f45fa5bdd3)
The User-Agent: header can be fun and interesting, but it also leaks
quite a bit of information about the user and their software stack.
This represents a potential security risk (attackers can target the
particular stack) and also an anonymity risk (a user trying to
preserve their anonymity by sending mail from a non-associated account
might reveal quite a lot of information if their choice of mail user
agent is exposed).
This change also avoids hiding the User-Agent header by default, so
that people who decide they want to send it will at least see it (and
can edit it if they want to) before sending.
It makes sense to have safer defaults.
Previously notmuch-wash made the width of the text (approximately) the
window-width minus the depth in thread. This is correct for the
default indentation of 1 per message depth, but is incorrect for any
other setting of notmuch-show-indent-messages-width.
As notmuch-show-indent-messages-width is customisable, and notmuch-tree
sets it to zero to avoid indenting messages in the message pane, this
bug can show up in real use.
Two of the tests had to be updated: when
notmuch-show-indent-messages-width is 0, then the new (correct) word
wrapping happens later, when notmuch-show-indent-messages-width is 4,
then the new word wrapping happens sooner.
gdb sometimes writes warnings to stdout, which we don't need/want, and
for some reason --batch-silent isn't enough to hide. So in this commit
we write them to a log file, which is probably better for debugging
anyway. To see an illustrative test failure before this change, run
% make
% touch notmuch-count.c
% cd test && ./T060-count.sh
This has been wrong since bbbdf0478e, but the race condition was not
previously been (often?) triggered in the tests. With the DB_RETRY_LOCK
patches, it manifests itself as a deadlock.
This support will be present only if the appropriate version of xapian
is available _and_ the user did not disable the feature when
building. So there really needs to be some way for the user to check.
This has been wrong since bbbdf0478e, but the race condition was not
previously been (often?) triggered in the tests. With the DB_RETRY_LOCK
patches, it manifests itself as a deadlock.
A few simple things that applies well to test/notmuch-test.sh
- Shell does pathname expansion also without doing `echo ...` in subshell.
- Redirections >/dev/null 2>/dev/null hide (improbable but) potential
serious errors; adding -f to rm instead.
- Inter-file capitalization consistency in comments.
- Unnecesary space removal.
The trick of having a common header file doesn't work to share between
test scripts, so make an include file in the test directory.
The use of #include <notmuch-test.h> looks slightly pretentious, but
the include file is not actually in the current (temporary) directory.
Place PYTHONPATH to the environment when python is executed in a way
that current shell environment is not affected. This also allows adding
the old value of PYTHONPATH to the end of the new value (otherwise it
would have been appended again and again when test_python is called).
At the same time, use -B option to avoid writing .pyc files to
bindings/python/* (which are not cleared out by distclean).
Drop the (unused) prefix code which preserved the original stdout of the
python program and opened sys.stdout to OUTPUT. In place of that there
is now note how (debug) information can be printed to original stdout.
Previously LD_LIBRARY_PATH was exported (and environment changed)
in the middle of test case execution, when a function setting it
was called.
Previously the old contents of LD_LIBRARY_PATH was lost (if any)
when it was re-set and exported. In some systems the old contents of
LD_LIBRARY_PATH was needed to e.g. locate suitable gmime library.
In scripts that include test-lib-common.sh but not test-lib.sh
the die() implementation needs to be a bit different due to
fd redirection differences. test-lib-common.sh implements die()
only if it was not implemented already.
Added die() function to test-lib.sh with the following first use of it:
If notmuch new fails during email corpus addition the database is
most probably inexistent or broken and the added corpus would be
unusable while running single tests, giving misleading failures
("only" full 'make test' cleans out old corpus).
Many of the external links found in the notmuch source can be resolved
using https instead of http. This changeset addresses as many as i
could find, without touching the e-mail corpus or expected outputs
found in tests.
Most of the infrastructure here is general, only the validation/dispatch
is hardcoded to a particular prefix.
A notable change in behaviour is that notmuch-config now opens the
database e.g. on every call to list, which fails with an error message
if the database doesn't exit yet.
Since xapian provides the ability to restrict the iterator to a given
prefix, we expose this ability to the user. Otherwise we mimic the other
iterator interfances in notmuch (e.g. tags.c).
This is a thin wrapper around the Xapian metadata API. The job of this
layer is to keep the config key value pairs from colliding with other
metadata by transparently prefixing the keys, along with the usual glue
to provide a C interface.
The split of _get_config into two functions is to allow returning of the
return value with different memory ownership semantics.
The Ruby bindings were missing a way to get all the tags of the
database. Now you should be able to access this with the public
instance method `all_tags` of your database object.
Example of use:
notmuchdb = Notmuch::Database.new path, { :create => false,
:mode => Notmuch::MODE_READ_ONLY }
my_tags = notmuchdb.all_tags
my_tags.each { |tag|
print tag
}
my_tags.destroy!
Amended by db: improve error reporting, add test
Files in test directories had only copyright of a single individual,
of which code was adapted here as a base of the test system.
Since then many Notmuch Developers have contributed to the test
system, which is now acknowledged with a constant string in some
of the test files.
The README file in test directory instructed new files contain a
copyright notice, but that has never been done (and it is also not
needed). To simplify things a bit (and lessen confusion) this
instruction is now removed.
As a side enchangement, all of the 3 entries in the whole source
tree cd'ing to `dirname` of "$0" now uses syntax cd "$(dirname "$0")".
This makes these particular lines work when current working directory
is e.g. /c/Program Files/notmuch/test/.
(Probably it would fail elsewhere, though.)
In case of notmuch reply and notmuch show --part=N it is required that
search terms match to one message. If match count was != 1, error
message "Error: search term did not match precisely one message"
was too vague to explain what happened.
By appending (matched <num> messages) to the error message it
makes the problem more understandable (e.g when <num> is '0'
user reckons the query had a typo in it).
To fully complete the ghost-on-removal-when-shared-thread-exists
proposal, we need to clear all ghost messages when the last active
message is removed from a thread.
Amended by db: Remove the last test of T530, as it no longer makes sense
if we are garbage collecting ghost messages.
implement ghost-on-removal, the solution to T590-thread-breakage.sh
that just adds a ghost message after removing each message.
It leaks information about whether we've ever seen a given message id,
but it's a fairly simple implementation.
Note that _resolve_message_id_to_thread_id already introduces new
message_ids to the database, so i think just searching for a given
message ID may introduce the same metadata leakage.
This test (T590-thread-breakage.sh) has known-broken subtests.
If you have a two-message thread where message "B" is in-reply-to "A",
notmuch rightly sees this as a single thread.
But if you:
* remove "A" from the message store
* run "notmuch new"
* add "A" back into the message store
* re-run "notmuch new"
Then notmuch sees the messages as distinct threads.
This happens because if you insert "B" initially (before anything is
known about "A"), then a "ghost message" gets added to the database in
reference to "A" that is in the same thread, which "A" takes over when
it appears.
But if "A" is subsequently removed, no ghost message is retained, so
when "A" appears, it is treated as a new thread.
I see a few options to fix this:
ghost-on-removal
----------------
We could unilaterally add a ghost upon message removal. This has a
few disadvantages: the message index would leak information about what
messages the user has ever been exposed to, and we also create a
perpetually-growing dataset -- the ghosts can never be removed.
ghost-on-removal-when-shared-thread-exists
------------------------------------------
We could add a ghost upon message removal iff there are other
non-ghost messages with the same thread ID.
We'd also need to remove all ghost messages that share a thread when
the last non-ghost message in that thread is removed.
This still has a bit of information leakage, though: the message index
would reveal that i've seen a newer message in a thread, even if i had
deleted it from my message store
track-dependencies
------------------
rather than a simple "ghost-message" we could store all the (A,B)
message-reference pairs internally, showing which messages A reference
which other messages B.
Then removal of message X would require deleting all message-reference
pairs (X,B), and only deleting a ghost message if no (A,X) reference
pair exists.
This requires modifying the database by adding a new and fairly weird
table that would need to be indexed by both columns. I don't know
whether xapian has nice ways to do that.
scan-dependencies
-----------------
Without modifying the database, we could do something less efficient.
Upon removal of message X, we could scan the headers of all non-ghost
messages that share a thread with X. If any of those messages refers
to X, we would add a ghost message. If none of them do, then we would
just drop X entirely from the table.
---------------------
One risk of attempted fixes to this problem is that we could fail to
remove the search term indexes entirely. This test contains
additional subtests to guard against that.
This test also ensures that the right number of ghost messages exist
in each situation; this will help us ensure we don't accumulate ghosts
indefinitely or leak too much information about what messages we've
seen or not seen, while still making it easy to reassemble threads
when messages come in out-of-order.
The code to skip multiple slashes in _notmuch_database_split_path()
skips back one character too much. This is compensated by a +1 in the
length parameter to the strndup() call. Mostly this works fine, but if
the path is to a file under a top level directory with one character
long name, the directory part is mistaken to be part of the file name
(slash == path in code). The returned directory name will be the empty
string and the basename will be the full path, breaking the indexing
logic in notmuch new.
Fix the multiple slash skipping to keep the slash variable pointing at
the last slash, and adjust strndup() accordingly.
The bug was introduced in
commit e890b0cf40
Author: Carl Worth <cworth@cworth.org>
Date: Sat Dec 19 13:20:26 2009 -0800
database: Store the parent ID for each directory document.
just a little over two months after the initial commit in the Notmuch
code history, making this the longest living bug in Notmuch to date.
In several places in the test suite we intentionally corrupt the Xapian
database in order to test error handling. This corruption is specific to
the on-disk organization of the database, and that changed with the
glass backend. We use the previously computed default backend to make
the tests adapt to changing names.
This is mainly for the test suite. We already expect the tests to be
run in the same environment as configure was run, at least to get the
name of the python interpreter. So we are not really imposing a new
restriction.
This should potentially be updated to have an equivalent optimization
for the glass backend, but it in my unscientific tests, the glass backend
without the optimization is faster then the chert backend with.
When no decryption or signature examination is
happening (i.e. `notmuch-crypto-process-mime' is `nil') insert buttons
that indicate this, rather than remaining silent.
in d27d90875d (2016-02-20) notmuch-mua-reply-insert-header-p-function
was set to notmuch-show-reply-insert-header-p-never as its default was
changed to something else. Now that default is set back to *-never so
this change done in d27d90875d is not needed anymore.
We only need a long string, not a single long term to trigger batch
mode. The giant term triggers a bug/incompatibility in Xapian 1.3.4
that throws an exception because it is longer than the Xapian term size
limit.
