RECOMMENDED is removed. It becomes ABI_DEPENDS.
BUILDLINK_RECOMMENDED.foo becomes BUILDLINK_ABI_DEPENDS.foo.
BUILDLINK_DEPENDS.foo becomes BUILDLINK_API_DEPENDS.foo.
BUILDLINK_DEPENDS does not change.
IGNORE_RECOMMENDED (which defaulted to "no") becomes USE_ABI_DEPENDS
which defaults to "yes".
Added to obsolete.mk checking for IGNORE_RECOMMENDED.
I did not manually go through and fix any aesthetic tab/spacing issues.
I have tested the above patch on DragonFly building and packaging
subversion and pkglint and their many dependencies.
I have also tested USE_ABI_DEPENDS=no on my NetBSD workstation (where I
have used IGNORE_RECOMMENDED for a long time). I have been an active user
of IGNORE_RECOMMENDED since it was available.
As suggested, I removed the documentation sentences suggesting bumping for
"security" issues.
As discussed on tech-pkg.
I will commit to revbump, pkglint, pkg_install, createbuildlink separately.
Note that if you use wip, it will fail! I will commit to pkgsrc-wip
later (within day).
and NetBSD-current which caused serius lossage:
depend on librfuncs>=1.0.7nb1 which implements NetBSD-current's
behaviour, change the patch to _gpgme_getenv() accordingly,
and bump PKGREVISION
following GNUPG recomendations: "Note that GnuPG 1.4 and 1.9 are not yet
in sync and thus features and bug fixes done in 1.4 are not available
in 1.9. *Please keep on using 1.4.x for OpenPGP*; 1.9.x and 1.4.x may
be installed simultaneously."
Noteworthy changes in version 1.0.2 (2004-12-28)
------------------------------------------------
* Changed the license of the library to the GNU Lesser General Public
License (LGPL), version 2.1 or later.
in the environment, we can handle this case properly from here. I.e., treat
ENOENT as if no error happened, and return NULL in the 'value' parameter (as
the function documentation says). Finally fixes PR pkg/27884.
Bump PKGREVISION to 2.
A test program was trying to run 'gpg', bypassing the binary name given
through the --with-gpg flag to configure.
XXX This still has a chance to break. If gnupg-devel is built w/o the
gpgsm option and gpgme is being built w/ it, no gpg binary will be found.
This is one of the those damages produced by build options...
2) Revised configure to perform header check for rfuncs.h
3) Revised files referencing getenv_r, strerror_r and ttyname_r
to include <rfuncs.h> if HAVE_RFUNCS_H
* Version 1.0.0! We are proud to present you with a thoroughly
tested and stable version of the GPGME library. A big Thank You!
to all the people who made this possible.
The development will be branched into a stable 1.x.y series and the
head.
* The gpgme.m4 macro supports checking the API version. Just prepend
it to the required version string, separated by a colon. For
example, this release has the version "1:1.0.0". The last release
to which this version is (mostly) ABI compatible is "1:0.4.2",
which is the default required version.
in the process. (More information on tech-pkg.)
Bump PKGREVISION and BUILDLINK_DEPENDS of all packages using libtool and
installing .la files.
Bump PKGREVISION (only) of all packages depending directly on the above
via a buildlink3 include.
All library names listed by *.la files no longer need to be listed
in the PLIST, e.g., instead of:
lib/libfoo.a
lib/libfoo.la
lib/libfoo.so
lib/libfoo.so.0
lib/libfoo.so.0.1
one simply needs:
lib/libfoo.la
and bsd.pkg.mk will automatically ensure that the additional library
names are listed in the installed package +CONTENTS file.
Also make LIBTOOLIZE_PLIST default to "yes".
Noteworthy changes in version 0.9.0 (unreleased)
------------------------------------------------
* The type gpgme_key_t has now a new field keylist_mode that contains
the keylist mode that was active at the time the key was retrieved.
* The type gpgme_decrypt_result_t has a new field "wrong_key_usage"
that contains a flag indicating that the key should not have been
used for encryption.
* Verifying a signature of a revoked key gives the correct result now
(GPG_ERR_CERT_REVOKED error code).
* Clarified that the error code GPG_ERR_NO_DATA from the decrypt &
verify operations still allows you to look at the signature
verification result.
* Clarified that patterns in keylisting operations have an upper
limit, and thus are not suited to list many keys at once by their
fingerprint. Also improve the error message if the pattern is too
long for the CMS protocol to handle.
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
gpgme_key_t EXTENDED: New field keylist_mode.
gpgme_decrypt_result_t EXTENDED: New field wrong_key_usage.
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
Set the 0.9.0 versions as BUILDLINK_DEPENDS in the buildlink3.mk for the
library major version changes.
Noteworthy changes in version 0.4.7 (2004-04-29)
------------------------------------------------
* Correctly initialize the fields expired, revoked, invalid, and
disabled in the gpgme_key_t structures.
* A bug fix: The flag wrong_key_usage of gpgme_signature_t was
accidently of type int instead unsigned int.
Noteworthy changes in version 0.4.6 (2004-04-06)
------------------------------------------------
* Bug fixes
Noteworthy changes in version 0.4.5 (2004-03-07)
------------------------------------------------
* GPGME is now compiled with LFS (large file support) by default.
* New key listing mode GPGME_KEYLIST_MODE_VALIDATE for validation of
the listed keys.
* New interface gpgme_cancel() that can be used to cancel
asynchronous operations.
Noteworthy changes in version 0.4.4 (2004-01-12)
------------------------------------------------
* The member "class" in gpgme_key_sig_t and gpgme_new_signature_t has
been renamed to "sig_class", to avoid clash with C++ compilers. In
the C API, the old name "class" has been preserved for backwards
compatibility, but is deprecated.
by moving the inclusion of buildlink3.mk files outside of the protected
region. This bug would be seen by users that have set PREFER_PKGSRC
or PREFER_NATIVE to non-default values.
BUILDLINK_PACKAGES should be ordered so that for any package in the
list, that package doesn't depend on any packages to the left of it
in the list. This ordering property is used to check for builtin
packages in the correct order. The problem was that including a
buildlink3.mk file for <pkg> correctly ensured that <pkg> was removed
from BUILDLINK_PACKAGES and appended to the end. However, since the
inclusion of any other buildlink3.mk files within that buildlink3.mk
was in a region that was protected against multiple inclusion, those
dependencies weren't also moved to the end of BUILDLINK_PACKAGES.
Noteworthy changes in version 0.4.3 (2003-10-06)
------------------------------------------------
* libgpgme should not be used for threaded programs anymore. This
never worked reliably in all cases, because you had to
be careful about the linking order and libtool wouldn't do that for
you automatically. Instead, now you have to link against
libgpgme-pthread for applications using pthread and libgpgme-pth for
applications using GNU Pth.
The old code for automagically detecting the thread library is
still part of libgpgme, but it is DEPRECATED.
* There are new automake macros AM_PATH_GPGME_PTH and
AM_PATH_GPGME_PTHREAD, which support checking for thread-enabled
versions of GPGME. They define GPGME_PTH_CFLAGS, GPGME_PTH_LIBS,
GPGME_PTHREAD_CFLAGS and GPGME_PTHREAD_LIBS respectively. These
variables of course also include the configuration for the thread
package itself. Alternatively, use libtool.
* gpgme_strerror_r as a thread safe variant of gpgme_strerror was
added.
* gpgme-config doesn't support setting the prefix or exec prefix
anymore. I don't think it ever worked correctly, and it seems to
be pointless.
* gpgme_get_key fails with GPG_ERR_AMBIGUOUS_NAME if the key ID
provided was not unique, instead returning the first matching key.
* gpgme_key_t and gpgme_subkey_t have a new field, can_authenticate,
that indicates if the key can be used for authentication.
* gpgme_signature_t's status field is now correctly set to an error
with error code GPG_ERR_NO_PUBKEY if public key is not found.
* gpgme_new_signature_t's class field is now an unsigned int, rather
than an unsigned long (the old class field is preserved for
backwards compatibility).
* A new function gpgme_set_locale() is provided to allow configuring
the locale for the crypto backend. This is necessary for text
terminals so that programs like the pinentry can be started with
the right locale settings for the terminal the application is running
on, in case the terminal has different settings than the system
default (for example, if it is a remote terminal). You are highly
recommended to call the following functions directly after
gpgme_check_version:
#include <locale.h>
setlocale (LC_ALL, "");
gpgme_set_locale (NULL, LC_CTYPE, setlocale (LC_CTYPE, NULL));
gpgme_set_locale (NULL, LC_MESSAGES, setlocale (LC_MESSAGES, NULL));
GPGME can not do this for you, as setlocale is not thread safe, and
there is no alternative.
* The signal action for SIGPIPE is now set to SIG_IGN by
gpgme_check_version, instead the first time a crypto engine is
started (which is not well defined).
* In the output of gpgme_hash_algo_name, change RMD160 to RIPEMD160,
TIGER to TIGER192, CRC32-RFC1510 to CRC32RFC1510, and CRC24-RFC2440
to CRC24RFC2440. For now, these strings can be used as the MIC
parameter for PGP/MIME (if appropriately modified).
Noteworthy changes in version 0.4.2 (2003-07-30)
------------------------------------------------
* Allow gpg-error to be in non-standard place when linking the test suite.
* Configure will fail now if gpg-error can not be found.
* Fixed initialized memory backed data objects for writing, which
caused the test program to crash (but only on Mac OS, surprisingly).
* Eliminate use of C99 constructs.
* Small improvements to the manual.
Noteworthy changes in version 0.4.1 (2003-06-06)
------------------------------------------------
This is the release that 0.4.0 should have been. There are many
interface changes, please see below for the details. The changes are
sometimes the result of new functionality, but more often express a
paradigm shift. Others are an overdue cleanup to get GPGME in line
with the GNU coding standards and to make the interface more
self-consistent. Here is an overview on the changes:
All types have been renamed to conform to the GNU coding standards,
most of the time by keeping the whole name in lowercase and inserting
underscores between words.
All operations consistently only accept input parameters in their
invocation function, and return only an error code directly. Further
information about the result of the operation has to be retrieved
afterwards by calling one of the result functions. This unifies the
synchronous and the asynchronous interface.
The error values have been completely replaced by a more
sophisticated model that allows GPGME to transparently and accurately
report all errors from the other GnuPG components, irregardless of
process boundaries. This is achieved by using the library
libgpg-errors, which is shared by all GnuPG components. This library
is now required for GPGME.
The results of all operations are now provided by pointers to C
structs rather than by XML structs or in other ways.
Objects which used to be opaque (for example a key) are now pointers
to accessible structs, so no accessor functions are necessary.
Backward compatibility is provided where it was possible without too
much effort and did not collide with the overall sanitization effort.
However, this is only for ease of transition. NO DEPRECATED FUNCTION
OR DATA TYPE IS CONSIDERED A PART OF THE API OR ABI AND WILL BE
DROPPED IN THE FUTURE WITHOUT CHANGING THE SONAME OF THE LIBRARY.
Recommendations how to replace deprecated or removed functionality
can be found within the description of each change.
What follows are all changes to the interface and behaviour of GPGME
in detail.
* If gpgme.h is included in sources compiled by GCC 3.1 or later,
deprecated attributes will warn about use of obsolete functions and
type definitions. You can suppress these warnings by passing
-Wno-deprecated-declarations to the gcc command.
* The following types have been renamed. The old types are still
available as aliases, but they are deprecated now:
[complete list in NEWS file]
* gpgme_error_t is now identical to gpg_error_t, the error type
provided by libgpg-error. More about using libgpg-error with GPGME
can be found in the manual. All error symbols have been removed!
* All functions and types in libgpg-error have been wrapped in GPGME.
The new types are gpgme_err_code_t and gpgme_err_source_t. The new
functions are gpgme_err_code, gpgme_err_source, gpgme_error,
gpgme_err_make, gpgme_error_from_errno, gpgme_err_make_from_errno,
gpgme_err_code_from_errno, gpgme_err_code_to_errno,
gpgme_strsource.
* GPGME_ATTR_IS_SECRET is not anymore representable as a string.
* GnuPG 1.2.2 is required. The progress callback is now also invoked
for encrypt, sign, encrypt-sign, decrypt, verify, and
decrypt-verify operations. For verify operations on detached
signatures, the progress callback is invoked for both the detached
signature and the plaintext message, though.
* gpgme_passphrase_cb_t has been changed to not provide a complete
description, but the UID hint, passphrase info and a flag
indicating if this is a repeated attempt individually, so the user
can compose his own description from this information.
The passphrase is not returned as a C string, but must be written
to a file descriptor directly. This allows for secure passphrase
entries.
The return type has been changed to gpgme_error_t value. This
allowed to remove the gpgme_cancel function; just return
the error code GPG_ERR_CANCELED in the passphrase callback directly.
* gpgme_edit_cb_t has been changed to take a file descriptor argument.
The user is expected to write the response to the file descriptor,
followed by a newline.
* The recipients interface has been removed. Instead, you use
NULL-terminated lists of keys for specifying the recipients of an
encryption operation. Use the new encryption flag
GPGME_ENCRYPT_ALWAYS_TRUST if you want to override the validity of
the keys (but note that in general this is not a good idea).
This change has been made to the prototypes of gpgme_op_encrypt,
gpgme_op_encrypt_start, gpgme_op_encrypt_sign and
gpgme_op_encrypt_sign_start.
The export interface has been changed to use pattern strings like
the keylist interface. Thus, new functions gpgme_op_export_ext and
gpgme_op_export_ext_start have been added as well. Now the
prototypes of gpgme_op_export_start and gpgme_op_export finally
make sense.
* gpgme_op_verify and gpgme_op_decrypt_verify don't return a status
summary anymore. Use gpgme_get_sig_status to retrieve the individual
stati.
* gpgme_io_cb_t changed from a void function to a function returning
a gpgme_error_t value. However, it will always return 0, so you
can safely ignore the return value.
* A new I/O callback event GPGME_EVENT_START has been added. The new
requirement is that you must wait until this event until you are
allowed to call the I/O callback handlers previously registered for
this context operation. Calling I/O callback functions for this
context operation before the start event happened is unsafe because
it can lead to race conditions in a multi-threaded environment.
* The idle function feature has been removed. It was not precisely
defined in a multi-threaded environment and is obsoleted by the
user I/O callback functions. If you still need a simple way to
call something while waiting on one or multiple asynchronous
operations to complete, don't set the HANG flag in gpgme_wait (note
that this will return to your program more often than the idle
function did).
* gpgme_wait can return NULL even if hang is true, if an error
occurs. In that case *status contains the error code.
* gpgme_get_engine_info was radically changed. Instead an XML
string, an info structure of the new type gpgme_engine_info_t is
returned. This makes it easier and more robust to evaluate the
information in an application.
* The new function gpgme_get_protocol_name can be used to convert a
gpgme_protocol_t value into a string.
* The status of a context operation is not checked anymore. Starting
a new operation will silently cancel the previous one. Calling a
function that requires you to have started an operation before without
doing so is undefined.
* The FPR argument to gpgme_op_genkey was removed. Instead, use the
gpgme_op_genkey_result function to retrieve a gpgme_genkey_result_t
pointer to a structure which contains the fingerprint. This also
works with gpgme_op_genkey_start. The structure also provides
other information about the generated keys.
* The new gpgme_op_import_result function provides detailed
information about the result of an import operation in
gpgme_import_result_t and gpgme_import_status_t objects.
Thus, the gpgme_op_import_ext variant is deprecated.
* The new gpgme_op_sign_result function provides detailed information
about the result of a signing operation in gpgme_sign_result_t,
gpgme_invalid_key_t and gpgme_new_signature_t objects.
* The new gpgme_op_encrypt_result function provides detailed
information about the result of an encryption operation in
a GpgmeEncryptResult object.
* The new gpgme_op_decrypt_result function provides detailed
information about the result of a decryption operation in
a GpgmeDecryptResult object.
* The new gpgme_op_verify_result function provides detailed
information about the result of an verify operation in
a GpgmeVerifyResult object. Because of this, the GPGME_SIG_STAT_*
values, gpgme_get_sig_status, gpgme_get_sig_ulong_attr,
gpgme_get_sig_string_attr and gpgme_get_sig_key are now deprecated,
and gpgme_get_notation is removed.
* GpgmeTrustItem objects have now directly accessible data, so the
gpgme_trust_item_get_string_attr and gpgme_trust_item_get_ulong_attr
accessor functions are deprecated. Also, reference counting is
available through gpgme_trust_item_ref and gpgme_trust_item_unref
(the gpgme_trust_item_release alias for the latter is deprecated).
* Keys are not cached internally anymore, so the force_update argument
to gpgme_get_key has been removed.
* GpgmeKey objects have now directly accessible data so the
gpgme_key_get_string_attr, gpgme_key_get_ulong_attr,
gpgme_key_sig_get_string_attr and gpgme_key_sig_get_ulong_attr
functions are deprecated. Also, gpgme_key_release is now
deprecated. The gpgme_key_get_as_xml function has been dropped.
* Because all interfaces using attributes are deprecated, the
GpgmeAttr data type is also deprecated.
* The new gpgme_op_keylist_result function provides detailed
information about the result of a key listing operation in
a GpgmeKeyListResult object.
* Now that each function comes with its own result retrieval
interface, the generic gpgme_get_op_info interface is not useful
anymore and dropped.
* The type and mode of data objects is not available anymore.
Noteworthy changes in version 0.4.0 (2002-12-23)
------------------------------------------------
* Key generation returns the fingerprint of the generated key.
* New convenience function gpgme_get_key.
* Supports signatures of user IDs in keys via the new
GPGME_KEYLIST_MODE_SIGS keylist mode and the
gpgme_key_sig_get_string_attr and gpgme_key_sig_get_ulong_attr
interfaces. The XML info about a key also includes the signatures
if available.
* New data object interface, which is more flexible and transparent.
Lots of interface changes, for details see the included
NEWS file.
Changes:
- The progress status is sent via the progress callbacks in
gpgme_op_edit.
- Bug fix for signing operations with explicit signer settings
for the CMS protocol.
From NEWS file:
Noteworthy changes in version 0.3.14 (2002-12-04)
-------------------------------------------------
* GPGME-Plug is now in its own package "cryptplug".
* Workaround for a setlocale problem. Fixed a segv related to not
correctly as closed marked file descriptors.
For full list of changes see ChangeLog.
(distfile checked against .sig file)
