* configure.ac (HAVE_GETTID,HAVE_SYS_GETTID): New test.
* src/debug.c: Include syscall.h if needed.
(tid_log_callback) [HAVE_SYS_GETTID]: Use SYS_gettid
--
Linux introduced the gettid syscall with 2.4.11 but glibc only with
its version 2.30. This patch allows building on older platforms.
Co-authored-by: lgh1
* autogen.sh: Update to version 2024-07-04 from libgpg-error.
* configure.ac (BUILD_COMMITID): New. Append to VERSION file.
* src/version.c (cright_blurb): Use BUILD_COMMITID here.
* src/conversion.c (spacep): New.
(_gpgme_strtokenize): New.
--
Function taken from GnuPG and license changed to LGPL 2.1. The
version in GnuPG was entirely written by the author.
src/encrypt-sign.c (encrypt_sym_status_handler): Add call of
_gpgme_encrypt_status_handler.
(encrypt_sign_start): Call _gpgme_op_encrypt_init_result with
success_required=1 if archive is created. Always call
_gpgme_op_sign_init_result with success_required=0 because the encrypt
status handler already checks for SUCCESS.
src/encrypt.c (op_data_t): Add success_seen flag.
(_gpgme_encrypt_status_handler): Return error if we didn't see a
required SUCCESS on GPGME_STATUS_EOF. Set success_seen flag on
GPGME_STATUS_SUCCESS.
(encrypt_sym_status_handler): Add call of _gpgme_encrypt_status_handler.
(_gpgme_op_encrypt_init_result): Add argument success_required. Set
success_seen flag if SUCCESS is not required.
(encrypt_start): Call _gpgme_op_encrypt_init_result with
success_required=1 if archive is created.
src/ops.h (_gpgme_op_sign_init_result, _gpgme_op_encrypt_init_result):
Add argument success_required to prototypes.
src/sign.c (op_data_t): Add success_seen flag.
(_gpgme_sign_status_handler): Return error if we didn't see a
required SUCCESS on GPGME_STATUS_EOF. Set success_seen flag on
GPGME_STATUS_SUCCESS.
(sign_init_result): Add argument success_required. Set success_seen
flag if SUCCESS is not required.
(_gpgme_op_sign_init_result): Add argument success_required and
forward it to sign_init_result.
(sign_start): Call sign_init_result with success_required=1 if archive
is created.
--
gpgtar emits a SUCCESS status just before successful termination. If the
process terminates unexpectedly (e.g. because it's killed) then gpgme
now reports GPG_ERR_EOF. The SUCCESS status is only required if a
signed and/or encrypted archive is created which is only supported for
OpenPGP. The other engines reject the GPGME_ENCRYPT_ARCHIVE flag so that
we don't need to check the protocol in the generic code.
This change also adds handling of invalid recipients in case symmetric
encryption is used which makes sense because one can combine symmetric
and public key encryption.
GnuPG-bug-id: 6554
* src/gpgme.h.in (GPGME_ENCRYPT_ADD_RECP, GPGME_ENCRYPT_CHG_RECP):
New flag values.
* src/engine-gpg.c (have_cmd_modify_recipients): New.
(gpg_encrypt): Check availability of the feature and prepare command.
* tests/run-encrypt.c (main): New options --add-recipients
and --change-recipients.
--
GnuPG-bug-id: 1825
* src/key.c (_gpgme_key_append_name): Support email-only user IDs with
upper case letters.
* tests/gpg/t-keylist.c (struct key_info_s): Add algo, length, sec_algo,
sec_length.
(keys): Add expected algo and length for primary and secondary subkeys.
(main): Factor out code for checking a key and the code for the keylist
test. Call the factored out test function and a new test function.
(check_key, test_keylist, key_with_email_only_user_id,
key_info_email_only_user_id,
test_email_only_user_id_with_upper_case_letters): New.
--
Email-only user IDs with upper case letters are now also parsed as a
user ID with empty name and the complete user ID as email.
GnuPG-bug-id: 7280
* src/version.c (do_subsystem_inits) [W32]: Switch gettext to UTF-8
for current thread and all new threads.
--
We did already enable UTF-8, but it was only effective for the main
thread. Now we enable it also for all new threads (if supported by
libgpg-error). This way debug output in background threads now also
uses UTF-8.
GnuPG-bug-id: 7188
* src/gpgme-tool.c (gt_get_keylist_mode): Increase NR_KEYLIST_MODES to
number of keylist modes. Add support for GPGME_KEYLIST_MODE_WITH_TOFU,
GPGME_KEYLIST_MODE_WITH_KEYGRIP, GPGME_KEYLIST_MODE_WITH_V5FPR. Write
all modes.
(hlp_keylist_mode): List all supported keylist modes.
(cmd_keylist_mode): Add support for GPGME_KEYLIST_MODE_WITH_TOFU,
GPGME_KEYLIST_MODE_WITH_KEYGRIP, GPGME_KEYLIST_MODE_WITH_V5FPR.
--
This fixes the possible overrun reported by Michal Hlavinka and adds
support for all keylist modes.
* src/Makefile.am (system_components): We don't use ath.h any more.
(main_sources): Remove ath.h and ath.c.
--
Signed-off-by: NIIBE Yutaka <gniibe@fsij.org>
* src/debug.c [HAVE_W32_SYSTEM or __linux] (tid_log_callback): Use
native thread ID call for Windows and Linux, ifdef-out-ed.
(debug_init): Only call gpgrt_log_set_pid_suffix_cb for relevant
systems.
--
Signed-off-by: NIIBE Yutaka <gniibe@fsij.org>
* src/b64dec.c: Remove.
* src/Makefile.am (main_sources): Remove b64dec.c.
* src/data-identify.c (pgp_binary_detection): Use gpgrt_b64dec in
libgpg-error.
--
gpgrt_b64dec is available in libgpg-error 1.27 and it is already in
use by JSON support.
GnuPG-bug-id: 7180
Signed-off-by: NIIBE Yutaka <gniibe@fsij.org>
* src/context.h (struct gpgme_context): New field import_options.
* src/engine-backend.h (struct engine_ops): Add arg import_options to
field 'import'.
* src/engine-gpg.c (gpg_import): Add arg import_options and pass option
--import-options with argument value to gpg.
* src/engine-gpgsm.c (gpgsm_import): Add dummy arg import_options.
* src/engine.c (_gpgme_engine_op_import): Add arg import_options and
pass option to import function of engine.
* src/engine.h (_gpgme_engine_op_import): Add arg import_options.
* src/gpgme.c (gpgme_release): Free 'import_options'.
(gpgme_set_ctx_flag, gpgme_get_ctx_flag): New flag "import-options".
* src/import.c (_gpgme_op_import_start, _gpgme_op_import_keys_start,
_gpgme_op_receive_keys_start): Pass import options stored in context to
_gpgme_engine_op_import.
* tests/run-import.c (show_usage, main): Add option --import-options.
--
This makes the --import-options option available in the GPGME API for
key imports.
GnuPG-bug-id: 7152
src/conversion.c, src/util.h (_gpgme_replace_backslashes): New.
src/engine-gpg.c (_add_arg): Add argument file_name. On Windows, replace
backslashes in file name arguments with forward slashes.
(add_arg_ext, add_arg_with_locp, add_arg, add_arg_pfx, add_gpg_arg,
add_gpg_arg_with_value, add_arg_len): Adjust call of _add_arg.
(add_file_name_arg, add_file_name_arg_len, add_gpg_arg_with_file_name):
New.
(add_file_name_arg_or_data): Call add_file_name_arg for file name.
(gpg_new): Call add_gpg_arg_with_file_name for home directory.
(gpg_decrypt): Call add_file_name_arg for output directory and output
file name.
(append_args_from_recipients_string): Call add_file_name_arg_len for
recipients file name.
(gpg_encrypt): Call add_file_name_arg for output file name, base
directory and input file name. Call add_gpg_arg_with_file_name to set
file name stored in message.
(gpg_encrypt_sign): Call add_file_name_arg for output file name, base
directory and input file name. Call add_gpg_arg_with_file_name to set
file name stored in message.
(gpg_sign): Call add_file_name_arg for output file name, base directory
and input file name. Call add_gpg_arg_with_file_name to set file name
stored in message.
(gpg_verify): Call add_file_name_arg for output directory and output
file name.
--
By passing all file names with forward slashes to gpg and gpgtar we
avoid problems caused by the quoting of backslashes.
GnuPG-bug-id: 7141
* src/engine-gpg.c: Add and use function for adding ADSKs.
* src/genkey.c: Prevent error due to no status line.
* src/gpgme.h.in: Add flag GPGME_CREATE_ADSK
* doc/gpgme.texi: Add documentation for ADSKs
* tests/run-genkey.c: Add test for adding ADSKs
--
This adds the ability to add ADSKs through
the gpgme_createsubkey interface.
The function must be called with NULL userid,
the ADSK fingerprint in algo
and the GPGME_CREATE_ADSK flag.
* src/engine-gpg.c (gpg_verify): Add "--verify" to command line.
--
Since GnuPG 2.1.16 --verify writes the signed data to the file specified
by --output. Explicitly telling gpg that we want to verify signed data
frees gpg from guessing what we want and avoids the corresponding
warning "no command supplied. Trying to guess what you mean ..."
GnuPG-bug-id: 6907
* src/engine-gpg.c (gpg_decrypt): Pass output file name to gpg if output
has file name set.
(gpg_verify): Ditto.
* tests/run-decrypt.c (show_usage): New option --output.
(main): Parse new option. Set file name on output if --output is given.
Do not print output if --output is given.
* tests/run-verify.c (show_usage): New option --output.
(main): Parse new option. Set file name on output if --output is given.
--
This change makes it possible to tell gpg to write the output (i.e. the
decrypted/verified data) directly to a file with given file name instead
of piping the output back to gpgme.
GnuPG-bug-id: 6550
* src/gpgme.h.in (GPGME_SIG_MODE_FILE): New signature mode flag.
* src/engine-gpg.c (gpg_sign): Separate signature mode from additional
flags. Check for incompatible flags. Explicitly set output to stdout if
no output file is used. Pass filename instead of fd to gpg when new flag
is set.
* src/engine-gpgsm.c (gpgsm_sign): Return error if new flag is set.
* src/engine-uiserver.c (uiserver_sign): Ditto.
* src/sign.c (sign_start): Consider new flag on check for invalid flags.
* tests/run-sign.c (show_usage): New options --detach and
--direct-file-io.
(main): Parse new options. Create a detached signature if --detach is
given. Make gpg read the input file itself if --direct-file-io is given.
--
With this change the gpgme_op_sign* functions gain the possibility to
make gpg read the data to sign directly from a file instead of from an
input FD to which it is written by gpgme.
GnuPG-bug-id: 6550
* src/gpgme.h.in (GPGME_ENCRYPT_FILE): New encryption flag.
* src/engine-gpg.c (gpg_encrypt, gpg_encrypt_sign): Check for
incompatible flags. Pass filename instead of fd to gpg when new flag is
set.
* src/engine-gpgsm.c (gpgsm_encrypt): Return error if new flag is set.
* src/engine-uiserver.c (uiserver_encrypt): Ditto.
* tests/run-encrypt.c (show_usage): New option --direct-file-io.
(main): Parse new option. Make gpg read the input file itself if the
option is given.
--
With this change the gpgme_op_encrypt* and gpgme_op_encrypt_sign*
functions gain the possibility to make gpg read the data to (sign and)
encrypt directly from a file instead of from an input FD to which it is
written by gpgme.
GnuPG-bug-id: 6550
* src/op-support.c (_gpgme_parse_plaintext): Decode filename
as percent string.
--
From gnupg/doc/DETAILS:
If a filename is available it gets printed as the third
argument, percent-escaped as usual.
so we can use the usual percent decode function here.
GnuPG-Bug-Id: T6852
* src/decrypt.c (_gpgme_decrypt_status_handler): Ignore received failure
code if we already have a specific failure code.
* src/encrypt.c (_gpgme_encrypt_status_handler): Ditto.
* src/export.c (export_status_handler): Ditto.
* src/genkey.c (genkey_status_handler): Ditto.
* src/keylist.c (keylist_status_handler): Ditto.
* src/keysign.c (keysign_status_handler): Ditto.
* src/passwd.c (passwd_status_handler): Ditto.
* src/revsig.c (revsig_status_handler): Ditto.
* src/setexpire.c (setexpire_status_handler): Ditto.
* src/sign.c (_gpgme_sign_status_handler): Ditto.
* src/tofupolicy.c (tofu_policy_status_handler): Ditto.
* src/verify.c (_gpgme_verify_status_handler): Ditto.
--
Usually, a process emits at most one failure code. But some operations
like the creation of an encrypted archive involve multiple chained
processes, so that multiple failure codes can be received. We want to
keep the first specific failure code we received. Further failure codes
are only parsed if we received just an unspecific "general error" so
far.
GnuPG-bug-id: 6575
* src/Makefile.am (gpgme_w32spawn_CFLAGS): Add -municode.
* src/gpgme-w32-spawn.c (build_commandline, my_spawn)
(translate_handles): Convert to wchar_t API.
(main): Use wmain instead.
--
Some time ago we introduced an inconsistency that w32-util called
gpgme-w32-spawn through CreateProcessW but since gpgme-w32-spawn
internally worked with 8 bit the chars were mangled and the
arguments not passed correctly through the CreateProcessA of the
child process. Since the GnuPG processes use GetCommandLineW
this is the proper way to pass on Unicode command line arguments.
Please note that we did not pass UTF-8 before this patch but
rather some broken native encoding where Windows replaces
unicode characters with question marks etc.
GnuPG-Bug-Id: T6728
* src/export.c (op_data_t): Add failure_code.
(export_status_handler): Set it.
(gpgme_op_export_ext, gpgme_op_export_keys): Also check the failure
code.
--
That is just in case we missed an error code. For example with older
gpg versions (e.g. 2.3.8) and exporting to a keyserver.
* src/gpgme.h.in (GPGME_KEYLIST_MODE_WITH_V5FPR): New.
(struct _gpgme_subkey): Add field v5fpr.
* src/engine-gpg.c (gpg_keylist_build_options): Pass new option to
gpg.
* src/key.c (gpgme_key_unref): Free new field.
* src/keylist.c (op_data_t): Parse and add "fp2" line.
* tests/run-keylist.c (show_usage): Add option --v5fpr.
* src/keylist.c (op_data_t): Add field failure_code.
(keylist_status_handler): Handle special value.
(gpgme_op_keylist_end): Return an error if a FAILURE line has been
seen.
--
Note that the failure code part has been added to better diagnose
problems if a wrong gpg version is used. If verything works right we
should not get this because we check that the gnupg version sis either
>= 2.4.4 or less than 2.3 and >= 2.2.42.
Note further that the v5fpr field may also be used to get the SHA-256
fingerprint of X.509 certificates (even without passing the new mode
flag).
GnuPG-bug-id: 6705
* src/engine-gpgsm.c (gpgsm_encrypt): Send the always-trust options.
* tests/run-encrypt.c: Add option --always-trust.
--
Note that the run-encrypt test tool used to assume always-trust for
OpenPGP since 1.7.0 This bug has also been fixed by introducing the
explicit option.
GnuPG-bug-id: 6559
* src/engine-gpg.c (gpg_verify): Use a separate pipe instead of stdout
for reading the plaintext.
* tests/gpg/t-support.h (PGM): Define if undefined.
(print_data): Undefine BUF_SIZE.
(check_data): New.
* tests/gpg/t-verify.c (clearsigned_plus_key_block): New.
(main): Add test.
--
Reading the plaintext from stdout is a bad idea because gpg can also
print other stuff on stdout, e.g. the keys contained in a public key
block. This is fixed by reading the plaintext via a special pipe.
GnuPG-bug-id: 6622
* src/verify.c (_gpgme_verify_status_handler): Return GPG_ERR_BAD_DATA
instead of GPG_ERR_GENERAL.
--
This gives users trying to decrypt and/or verify for example a public
key block a more useful error.
GnuPG-bug-id: 6617
