2019-02-07 iulius * Update to latest upstream config.guess, config.sub and libtool versions * Bump revision number for next release and copyright year 2019-02-04 iulius * Allow again the use of UTF-8 in header fields Regression since INN 2.6.1. Posts with internationalized header fields (UTF-8) are now accepted again. 2018-12-29 iulius * Update libtool from upstream 2.4.6-6 Debian version * Update to latest upstream config.guess and config.sub files * Update to latest pgpverify 1.30 The previous INN release was done before the official release of pgpverify 1.30; that's why a few comments differed. * Bump version information for our libraries released with INN 2.6.3 A few interfaces have changed in libinn (fqdn function, and new innconf parameter). * nnrpd: improve logging of failed connections Properly logs the hostname of clients whose connection failed owing to an issue during the negotiation of a TLS session or high load average. Nothing was previously logged. Thanks to Paolo Amoroso for the bug report. 2018-12-17 iulius * Fix a const discarding when assigning a variable Python 3.7 changed the return value of PyUnicode_AsUTF8 from char to const char. * Update to latest C TAP Harness upstream version 2018-12-02 iulius * Update comments of TLS implementation (it is not v1 only) * Take into account return values of chdir and system commands * Update to latest rra-c-util upstream version, improving Perl and Python support at configure time * Update to latest rra-c-util upstream version The Autoconf m4 macros now check that the library's key header can be found as well as the library. It removes the need of several checks added in configure.ac to achieve that. Also fix INN_LIB_KRB5_OPTIONAL probing in the --enable-reduced-depends case when linking with libkrb5 is not possible but libcom_err is present. The probes for com_err were previously not skipped, resulting in a non-empty KRB5_LIBS containing only -lcom_err, which incorrectly caused the macro to indicate that Kerberos libraries were present. * Take into account new GCC 7.0 and 8.0 warnings Merge commits 10193 and 10291 from CURRENT. * Rewrite GetFQDN and rename to inn_getfqdn Use getaddrinfo with AI_CANONNAME instead of the tedious code to find an FQDN in gethostbyname results. Return a newly-allocated string instead of using a static buffer, since this function isn't performance-critical, avoiding buffer management complexity. Adjust call sites to do proper memory allocation handling of the return value. Remove some old disabled code to try to force canonicalization by adding a period to the end of the DNS query. Patch from Russ Allbery (backported from CURRENT). 2018-11-11 iulius * Mention in POD documentation that TLS 1.3 no longer allows TLS-level compression * inn.conf: Add new tlsciphers13 parameter to fine-tune TLS 1.3 cipher suites A separate cipher suite configuration parameter is needed for TLS 1.3 as TLS 1.3 cipher suites are not compatible with TLS 1.2, and vice-versa. The tlsciphers13 parameter is based on the already existing tlsciphers parameter for TLS 1.2 and below. 2018-11-10 iulius * Fix the use of elliptic curve selection OpenSSL 1.1.0 removed the SSL_CTX_set_ecdh_auto function introduced with OpenSSL 1.0.2. This removal caused INN to always enforce NIST P-256 instead of using the most secure curve (which OpenSSL 1.1.0 does by default). Now fixed. 2018-06-03 eagle * Enable Perl and Python when testing snapshots 2018-05-21 iulius * Update to latest C TAP Harness and rra-c-util upstream versions * Add trailing comment in CHECK and TAKETHIS responses Responses to CHECK and TAKETHIS commands did not mention the reason of a reject (notably from Python or Perl filter hook, but also directly from innd main checks). This is now fixed. closes #148 2018-05-14 iulius * Various number of typo fixes Thanks to Christoph Biedl for the bug report. * Add support for embedded Python 3 interpreter to use with innd and nnrpd filter hooks * Update configure script to find Python 3 interpreter and correctly set linker flags for embedding it. Python 3.3.0 or later in the 3.x series is now supported. * Drop support for Python 2.2.0; now, Python 2.3.0 or later in the 2.x series is required because configure looks for a shared Python library, installed in the main library location in Python 2.3.0 and later. * Add m4 macros to check for minimal Python version and module presence at configure time. * Notable changes for Python 3 are: - string literals are now considered as Unicode data whereas they were mere bytes in Python 2. Consequently, encoding now really matters for strings, and UTF-8 is required in return values of Python filter hooks. (Note that compliance with NNTP would also want strings to be encoded in UTF-8.) Care should be given to use the right str or bytes objects in Python 3. - buffer objects no longer exist: they have been replaced with memoryview objects. Consequently, code to deal with them slighty changed. - integers no longer exist in the C API: they have been replaced with long integers (Py_ssize_t). - embedded Python initialization is now done differently (with a PyMODINIT_FUNC function). Try to homogeneize initialization with Python 2.x. * Improve error/notice logs. * Update and improve both documentation and samples accordingly. * Remove unused PYpathkey char * variable. * A few typo fixes. * Fix prototypes in clientlib man page * Typo fixes 2018-03-22 iulius * Bump revision number to next 2.6.3 version for STABLE