* lang/python/examples/howto/export-secret-keys.py and groups.py:
Updated the backwards compatibility adjustments to account for
unicode differences between python 2 and 3.
* lang/python/examples/howto/groups.py: subprocess update
* lang/python/examples/howto/export-secret-keys.py: subprocess update
Both of these try the nice and easy method of getting the subprocess
output available in Python 3, but will fall back to the older Popen
method if it doesn't work. Essentially this is to be a little nicer
to Python 2.7.15 (even though the examples are filled with warnings
that py2 support is not guaranteed with the examples).
--
* src/Helpers.js: GPGME_Keys were not parsed as valid, as their
fingerprint getter is not a fingerprint 'property'.
* BrowserTestExtension: fixed a dsplay typo in counting of tests.
--
* BrowsertestExtension/tests/decryptTest.js: There were cases in which
file names returned in a wrong encoding from decryption. The test
cases here are a 'Hello World' in a text file with different names,
then being encrypted with cli gnupg.
--
* src/Helpers.js: This additional escape should 'repair' special
characters like spaces in filenames. In the strange world of
encoding there is little hope that this captures all cases, or
that it will never fail to return some value, let alone meaningful.
In my test cases it worked.
--
* BrowserTestExtension/tests:
- decryptTest.js: Check Decryption and return values of binary data
- encryptTest.js: Return data type of armored/non-armored encryption
- added a small encoded input png for testing
* DemoExtension/maindemo.js: Fixed unexpected usage of the Demo encrypt
(non-armored)
--
* src/gpgme.js: In case the encryption was done unarmored, the result
is binary data. Added an option to either return the binary data as
base64-encoded string or as Uint8Array, similar to return values of
decrypt
--
* src/Connection.js; src/permittedOperations.js: To avoid further
encoding problems, data sent by gpgme is now sorted as either
'payload' or 'info'. Payload data may come in any encoding, and here
the 'expected' and 'format' options are used, 'info' data may
contain text created by gnupg which may need re-encoding, but this
should not be affected by 'expected' and 'format'
* lang/python/src/core.py: Adjusted new_from_estream function to alias
new_from_stream instead of fd.
* fixed the _gpgme import errors introduced in commit
08cd34afb7 by changing the exported
functions/types to match the inner module where all the work is
done, rather than the outer one(s).
Tested-by: Ben McGinnes <ben@adversary.org>
Signed-off-by: Ben McGinnes <ben@adversary.org>
--
* src/gpgmejs.js/encrypt: the encrypted data were converted back to a
(incorrect) string, whereas they should be data with no encoding
specified. Returning base64 data is the expected way.
* DemoExtension: caught yet another usage of old syntax.
* lang/python/docs/GPGMEpythonHOWTOen.org: Updated links to the
ProtonMail keyserver import scripts and added a warning regarding
being unable to update third party keys.
* lang/python/examples/howto/pmkey-import-alt.py: added usage.
* lang/python/examples/howto/pmkey-import.py: added usage.
--
* src/Helpers.js: As non-payload data might come in different
encodings, a conversion has been introduced that worked in most
cases. Data like the userid might come in different encodings,
which we don't know of. For now, a try..catch returns the data
as they are if the utf-8 decoding fails. Sometimes this yields the
correct result, sometimes it may not work, but it won't stop the
whole operation anymore.
--
* destructuring just takes the input argument and treats it as object.
In cases like in src/Keyring/generateKey, where I forgot to change
the old syntax, the fingerprint as string was destructured into an
object without "pattern", which caused all Keys to be retrieved.
So, methods with a destructuring now check if the first argument is
an object and get a default empty object if no parameter is
submitted. This allows the further use of destructured parameters,
while still ensuring nothing vastly incorrect is used.
* src/Kering.js, unittsets.js: fixed old syntax in method usage
--
* src/Connection.js: resulting data, if not pure ascii, is base64
encoded in the result message. A further decoding attempt into
javascript 'string' will be attempted by default, unless specified
at the decrypt() method. The return value 'format' now shows which
of the possibilities has been applied. The old boolean 'base64'
now turns into format:'base64' if the returned payload is a base64
string after decryption.
--
* reflecting the new optional strings accepted by the backend.
'file_name' and 'sender' can be used via the 'additional'
parameter in encrypt operations
--
* recent changes in parameter calling led to a forgotten internal call
in getDefaultKey using old syntax (and failing in case a default key
is configured)
--
* src/gpgmejs.js: Setting the default to 'always trust' assumes that
most api users will already have made their internal checks, but may
not have the gnupg web-of-trust model implemented, thus trusting the
key themselves, without gnupg having full or even any information.
Still it should stay an option to have gnupg decide.
--
* src/Keyring.js: Adapted Keyring.getDefaultKey() to my current
understanding of a default signing key: either the default key set
in the gpg config, or 'the first usable private key' - usability
meaning 'not invalid, expired, revoked, and can be used for
signing'. It should be the same key used as in command line when
doing a --sign operation.
In case the user has a smartcard plugged in, we currently
won't know of this here, so our choice may differ. But as we do all
javascript-binding sign operations with the key fingerprint
explicitly set, this should not be a real problem. This method is
seen more as a convenience to tell using librarys which key
represents the main user.
--
* As a decrypt result cannot be known beforehand, the decrypt operation
may add an 'expect' property, taking either 'uint8' or 'base64',
which will return the decrypted data in the appropiate formats.
the return property 'format' will give a feedback on which option
was taken.
A test was added to reflect these changes.
--
* As requested by using parties, the options to be passed into the
methods are now objects, with the objects' properties better
describing what they do, and to avoid the need to type several nulls
in a method call if one wants the last parameter.
- src/Keyring.js, src/gpgme.js: Changed parameters and their
validations
- BrowserTest/*.js Had to adapt quite some calls to the new format
--
* src/Connection.js, src/Helpers.js: performance of decoding incoming
base64 data was improved to about 4 times the speed by introducing
two more efficient functions (thanks to rrenkert@intevation.de for
finding and testing them)
* src/gpgmejs.js: Decrypted data will now return as Uint8Array, if the
caller does not wish for a decoding. Decoding binary data will return
invalid data, and a Uint8Array may be desired. This can be indicated
by using the (new) 'binary' option in decrypt.
* src/Errors.js A new error in case this decoding fails
* src/Message.js, src/Connection.js: expected is change from base64
to binary, to avoid confusion later on.
--
* src/Signature.js/get fingerprint: A signature with no fingerprint
should not happen, but if it does, we should throw an error here,
as the method is a getter.
This adds a new language binding "gpgme.js" to GPGME. It
serves as a bridge between the native-messaging service "gpgme-json"
and JavaScript Applications.
The first user of this binding will be Mailvelope which will
see GnuPG integration in the near future.
GnuPG-Bug-Id: T4107
--
* synchronous functions should throw errors if something goes wrong,
Promises should reject. This commit changes some error cases that
returned Error objects instead of throwing them
- src/Key.js: createKey() and sync Key.get() throw errors
- src/Error.js: Exporting the list of errors to be able to test and
compare against these strings
- src/Keyring.js: Setting a null value in pattern is not useful, and
now caused an error with the new changes.
- src/Message.js: createMessage and Message.setParameter now throw
errors
--
* src/gpgmejs.js: Decrypt now parses additional optional dec_info
information, as well as any verify information, if present
* src/permittedOperations: Now decrypt also expect the new return
object dec_inf (containing info such as is_mime and file_name)
--
* src/Keyring.js: Changed key ecpiration from Date to seconds from
creation, as in gpgme. The Date parameter used before was due to a
misunderstanding in documentation and requests from potential users.
--
* undoes 94ee0988d4 and
e16a87e839.
I do not fully understand why my approach was bad, but I am not in
a position to argue. This revert was requested to me after a review,
and I'm doing it in the assumption that more experienced people know
better than me.
* unittests: Also changed some outdated tests that stopped working
since 754e799d35 (as GPGME_Key is not
exported, one cannot check for instanceof in the tests anymore)
* import-key.py: fixed a minor typo.
* pmkey-import.py: locates and imports keys from the ProtonMail keyserver.
* pmkey-import-alt.py: the same as the previous except with setting an
alternative $GNUPGHOME directory.
* Moved the build import back up where it belongs.
* Included comments indicating how to build and install for multiple
Python versions beyond the first 2 on the same system.
* lang/python/version.py.in: Fixed most things, but there's still an
issue near the build portion with the existing Python bugs referenced.
* lang/python/setup.py.in: Now PEP8 compliant.
* PEP8 compliance for all constants except the globals in
src/constants/__init__.py depending on whether the import sequence
affects the globals themselves.
--
* Arriving strings (i.e. user id names, error messages) are not
always in javascript encoding. This is an attempt to go through
the whole gpgme answer (with the exception of payload data) and
to fix the encoding of these
--
* src/Key.js The synchronous mode for a Key does not offer an armor/
armored property anymore. This frees up a lot of performance issues,
also the armored expoort is expected to change quite often, so a
cached version is not advisable.
* hasSecret/getHasSecret is now refactored, to reflect their uses.
With get('hasSecret') there is a method that fetches the result.
* src/Key.js also some refactoring
--
* src/Key.js Key class is not exported anymore, as it should not be
used directly anywhere. setKeyData is no more a method of the Key,
(optional) data are now validated and set on Key creation and on
updates, both from within this module, thus no longer exposing
setKeyData to the outside.
* createKey now gained an optional parameter which allows to set Key
data at this point.
--
* src/Keyring.js: An empty result should no longer cause an error,
the import feedback summary has been refactored slightly
* Browsertests to reflect import feedback change
--
* src/Keyring.js I wrongly assumed an object to be a GPGME_Key,
it was the raw answer from nativeMessaging instead. Now it returns
a GPGME_Key again.
--
* src/Keyring.js: The answer was not parsed correctly, so a config was
being ignored.
* If no config is set, we return the first non-invalid key with a
secret, instead of the first key (which may be e.g. an expired one)
--
* src/Keyring.js: In case no default Key is set in configuration,
only Keys reported as having a secret part should be considered
for default Keys, avoiding some extra requests.
--
* src/Key.js: the fingerprint returned by a Key is now always upper
case hex, even if the constructor had lower case input. This is to be
more consistent with gpgme and to be more readable and reliable in
comparisions.
* lang/python/examples/howto/symcrypt-file.py: *sigh*; passphrase was
right the first time, just the error check that wasn't.
* I really should stop second guessing myself one of these days ...
Signed-off-by: Ben McGinnes <ben@adversary.org>
* Ran all the .py files in src/ and below through Yapf.
* Included some manual edits of core.py, this time successfully making
two notorious sections a bit more pythonic than scheming.
* Left the module imports as is.
* This will be committed if it passes the most essential test:
compiling, installing and running it.
Signed-off-by: Ben McGinnes <ben@adversary.org>
* lang/cpp/src/gpggencardkeyinteractor.cpp
(GpgGenCardKeyInteractor::Private::keysize): Change to string.
--
The value is only required as string so we can save it this
way to avoid the need to convert it for the action command.
GnuPG-Bug-Id: T4094
* lang/python/tests/Makefile.am,
lang/qt/tests/Makefile.am,
tests/Makefile.am,
tests/gpg/Makefile.am,
tests/gpgsm/Makefile.am,
tests/opassuan/Makefile.am (GNUPGHOME): Make variable explict.
--
If the build directory has too long path, gpgme could fail.
This is similar to
https://bugs.debian.org/cgi-bin/bugreport.cgi?bug=847206
In order to fix that, this patch extracts the GNUPGHOME variable
to be presented directly in the Makefile and thus overridable by
command line option.
A build system can then create a symlink to the GNUPGHOME directory
in /tmp and use that symlink as the GNUPGHOME directory
thus making the path very short.
GnuPG-Bug-Id: T4091
Patch provided by vlmarek
--
* src/index.js: The export now uses a freezed Object, which does not
allow for simply overwriting the init method by e.g. a third-party
library.
* BrowsertestExtension: Added some tests trying if decryption of bad
data properly fails
--
* The aim is to iterate through the results of the first request
(all keys), and then add the propert 'hasSecret' to those that
are in the second request (secret Keysring) as well. I messed
this up in a recent change, and it escaped testing.
--
* An Object.freeze should stop any malicious third party from changing
objects' methods once the objects are instantiated (see unittest for
an approach that would have worked before)
- An initialized gpgmejs- object doesn't have a '_Keyring' property
anymore (it still has its 'Keyring')
- The internal expect='base64' needed to be turned into a method.
--
* src/ [Connection, Error, Key, Keyring, MEssage, Signature, gpgmejs]:
Functions and values that are not meant to be overwritten are now
moved into their constructors, thus eliminating the possibility of
overwrites after initialization.
* Key: The mode of use (synchronous cached, or async promises) ivs now
determined at initialization of that Key. The property Key.isAsync
reflects this state.
* unittests: fixed old Key syntax for testing.
* Message.js isComplete is now a method and not a getter anymore.
* Added some startup tests.
--
Tests will now run with one instance of gpgmejs each block,
which reduces overhead. Readability is (hopefully) improved),
some negative tests are added.
There is still a performance problem in base64 encoding/decoding,
which causes some tests to fail due to time out.
--
* The import answer now also directly contains the armored Key as Key
property, without need to refresh the Key object created in the
answer. This allows for direct comparision of input and output.
* BrowserTestExtension: added test for that import callback
--
* The libgpg error strings arrive in the browser in a different
encoding than used by browsers. Escaping and then decoding it
should cover most languages in the supported browsers.
* lang/cpp/src/key.cpp (Key::update): Check that the key is
not NULL.
* lang/cpp/src/verificationresult.cpp (GpgME::Signature::key):
Check for fingerprint.
* Fixed and tested the changes necessary for org-mode to correctly
parse pythonic (Python 3) indentation.
* Updated the source blocks to recommended upper case for BEGIN_SRC
and END_SRC.
* Tested and confirmed XHTML output matches correct examples.
* Tested against pseudo-control output via exporting from org-mode to
org-mode and then exporting that to XHTML. Remaining differences
appear to be discarding the custom tags used to provide X[HT]ML id
elements to each section which does not appear to offer any benefit.
* Exporting directly to XHTML or other HTML output should no longer
cause problems, but if there are any then the first step should be
exporting from org-to-org and then exporting that to XHTML.
Tested-by: Ben McGinnes <ben@adversary.org>
Signed-off-by: Ben McGinnes <ben@adversary.org>
* Another retrofitting of the HOWTO Python example code, this time
following adjustments to python-mode configuration and having
trawled through the org-mode mailing lists for clues.
* lang/qt/src/threadedjobmixin.cpp (fromEncoding)
(stringFromGpgOutput): New helpers.
(markupDiagnostics): Use it.
--
The Problem is that on my western windows system GnuPG
gets CP 437 as GetConsoleOutputCP and prints in that codepage.
In a W32 GUI Application we get 0 as GetConsoleOutputCP and 1252
with GetACP.
The only thing that seemed to somehow match was GetOEMCP but
that might just be luck and it might still be broken in
other windows languages.
This code is also used in Kleopatra so it might make sense
to make it public once it is demonstrated that it works on
most systems.
--
* src/Keyring.js: If the optional "pattern" parameter is not to be
used, but another, following parameter is, null is more of a
convention in javascript, thus both null and undefined are
interpreted as "this parameter is not meant to be set".
