---------------------------------------------------------------------- Version 1.12.0 - 2025-09-08 - API: add ability to communicate source port back to caller This is handled via a new "on connection" callback - tcp: tcp: avoid deprecated ENGINE_cleanup ENGINE_cleanup is deprecated as of OpenSSL 1.1.0, and may not be available with 3.0 and OPENSSL_NO_ENGINE. Thanks to Yaakov Selkowitz for the patch. - relp: fix build against `gcc-14` (`-Werror=calloc-transposed-args`) `gcc-14` added a new `-Wcalloc-transposed-args` warning. It detected minor infelicity in `calloc()` API usage Fixes ../../git/src/relp.c: In function 'addToEpollSet': ../../git/src/relp.c:101:39: error: 'calloc' sizes specified with 'sizeof' in the earlier argument and not in the later argument [-Werror=calloc-transposed-args] 101 | CHKmalloc(epd = calloc(sizeof(epolld_t), 1)); Thanks to Khem Raj for the patch. - tests: Include missing sys/time.h This is found when building for musl C library systems where sys/time.h is not included indirectly and select() and timeval structs are used Thanks to Khem Raj for the patch. - bugfix: code adapted to C23 (as used in gcc 15 by default) Thanks to Khem Raj for the patch. ---------------------------------------------------------------------- Version 1.11.0 - 2023-01-10 - bugfix: function inline errors in debug optimization (-Og) Compiler does not inline any functions when using debug optimization (-Og). Hence, remove -Winline flag when compiling with debug optimization. Thanks to Yash Shinde for the patch. - CI: upgrading to newer ubuntu version as required by github we used older ubuntu versions as base for our containers, these versions are no longer available in github actions. As such, we upgraded. Has no effect on testbench as we execute everything in container environment and container remains the same. - tcp: fix some compiler warnings with enable-tls-openssl When --enable-tls=no and --enable-tls-openssl=yes, the following compiler errors were reported. Thanks to Patrick Williams for the patch. - code cleanup During the initial librelp implementation it was thought about an auto-retry feature (whatever that means). Variable bAutoRetry was created to control that behaviour. It's default was 0 (inactive). Some code has been written to evaluate that variable and act accordingly. However, no setter method for this variable was created and it was also nowhere changed from the default value of zero. As such, the code activated by it was never used. This patch now removes both the variable and the few code lines accessing it. This will have no effect, as the code was never used. closes: https://github.com/rsyslog/librelp/issues/223 - testbench: Fix commands in some openssl tests Those commands were not correctly handeled on AIX, now they are. - AIX: Changed ERRNO handling after connect in tcp.c see also: https://github.com/rsyslog/librelp/issues/250 - AIX: Add handling for other ERRNO codes in tcp.c For send(), recv() and connect() we need to handle different ERRNO codes on AIX according to doc: https://www.ibm.com/docs/en/aix/7.1?topic=r-recv-subroutine https://www.ibm.com/docs/en/aix/7.1?topic=s-send-subroutine https://www.ibm.com/docs/en/aix/7.1?topic=c-connect-subroutine - bugfix/TCP: relpTcpGetRtryDirection onyl needs to check direction if SSL is active. closes: https://github.com/rsyslog/librelp/issues/240 - AIX: in relpTcpRcv we need to set RETRY_recv if errno is 0 On AIX 10.x, when a recv call fails with len -1 and errno 0 we can still go into relpTCP_RETRY_recv mode. Testbench on rsyslog side appears to be finally working with this fix. On AIX 10.x, same problem as above appear to happen with connect() as well not setting errno to EINPROGRESS on failure. rsyslog ptcp driver does not check errno value at all, which is why we did not have this problem on AIX with rsyslog. Reduce test message count in receiver-abort.sh to 50000, because otherwise the test can timeout on very slow machines. Moved receive buffer in relpSessRcvData back to stack memory closes: https://github.com/rsyslog/librelp/pull/239 - openssl: fix openssl exit code avoid double free of ctx When more than one librelp instance are used in the same process, and the relpTcpExitTLS call was called a second time, the process could freeze due a double free (See SSL_CTX_free call). Thanks to David GOUARIN for the discovery and initial fix. Also added an openssl instance counter to avoid freeing the SSL library which could be still in use by another librelp instance in the same process. closes: https://github.com/rsyslog/librelp/issues/235 - librelp hardening: Fix multiple minor issues causing debugging trouble - avoid invalid dbgprint calls - avoid double free in relpTcpDestruct (if called twice). - add debug output into relpTcpRcv - OpenSSL: fix depreacted API issues for OpenSSL 3.x - OpenSSL error strings are loaded automatically now - Adapted Debug Callback for 3.x - See for more: https://www.openssl.org/docs/manmaster/man7/migration_guide.html closes: https://github.com/rsyslog/librelp/issues/246 - bugfix: compatiblity problem with openssl 1.1 There were undefined reference to CRYPTO_set_id_callback if system use openssl 1.1, which lead to load and/or other erros. Thanks to Alexander Eremin for the patch. - bugfix: Forward return code from relpEngineSetTLSLib to relpEngineSetTLSLibName Previous situation: In case librelp is compiled without openssl support but openssl is requested from the user side, then librelp silently switches to the default tls driver(gnutls). Thanks to Attila Lakatos for the patch. - bugfix: make relpEngineSetTLSLib debug safe In case pThis == NULL condition is true at line https://github.com/rsyslog/librelp/blob/master/src/relp.c#L345 pThis->dbgprint(...) would cause unexpected behavior. Thanks to Attila Lakatos for the patch. - bugfix: warnings reported by coverity scan Thanks to Attila Lakatos for the patch. - testbench: Added compatiblity to run on AIX - removed long options from receive send utilities (Not supported on AIX). - added checks for existing commands in some tests. - added msleep utility from rsyslog (sleep, timeout etc. not supported on AIX). - github workflow: Disabled some unrelated tests from rsyslog testbench closes: https://github.com/rsyslog/librelp/issues/228 - gnutls drvr bugfix: library called exit() under some circumstances When relpTcpRcv ran into an error, it could call exit() and thus terminate the calling process. This was a regression from commit 6555545c4ebe9afe9259b15c176fd81eeecc93f4. closes: https://github.com/rsyslog/librelp/issues/230 ---------------------------------------------------------------------- Version 1.10.0 - 2021-02-16 - TLS handling bugfix When the client tried to conntect to the server, custom cipherstrings (Set by tlscommands feature) were not used. This could lead to the negotiation of different and potentially weaker ciphers. Other custom tlscommands settings like Protocol where not affected. We do not overwrite the custom ciphers anymore if they are set by tlscommands. Another problem only related to the relp receiver (server) was, that the custom tlscommands/priority string where not applied on the accepted client connections. This could lead to the same problem as the default ciphers were used. Besides the main problem, the following changes were applied: - Add new testcase for setting custom tls ciphers in tlscommand. - Add support to use semicolon (;) as tlscommand seperator (See new testcase) - Fix GCC9 "destination size" warning in offers.c closes: https://github.com/rsyslog/librelp/issues/224 ---------------------------------------------------------------------- Version 1.9.0 - 2020-11-24 - openssl bugfix: preprocessor check for tlsconfigcmd code tls config commands no is compiled for OpenSSL 1.0.2. closes: https://github.com/rsyslog/librelp/issues/221 - solaris compatibility fix: add strndup compatibility code On some platforms, strndup() is not present. However, it was used in "tlsconfigcmd" command processing. This bug was unlikely to trigger as the platforms in question had either non-appropriate openssl library versions or were affected by the openssl version check bug (see above). ---------------------------------------------------------------------- Version 1.8.0 - 2020-09-29 - gnutls "bugfix": handle receives who break connection on close Some TLS servers don't reply to graceful shutdown requests "for optimization". This can lead to librelp keeping the connection for ever and thus effectively to hang on it. This patch is modelled after a similiar patch by Renaud Métrich for rsyslog. see also https://github.com/rsyslog/rsyslog/pull/4424 - gnutls bugfix: per-session memory leak Depending on callig sequence and use, there were two memory leaks which could cause operational issues for long-running processes. This was primarily a problem for sender processes (and not listeners) closes: https://github.com/rsyslog/librelp/issues/194 - tls bugfix: RETRY not correctly handled in TLS Mode & CI improvement Relp session handling did not honor rtryOP status correctly, instead it was checkong for errno=EAGAIN only. This patch comes with a number of smaller patches, unfortunately all rolled into a single commit. - CI: Add centos8 / fedora30 workflows. - CI: Switch rsyslog workflow to ubuntu-20.04 - CI: Imported run-ci script from rsyslog - Fix multiple issue found with -Werror=missing-declarations (fedora30). - openssl: Fix error output for all error cases - testbench: Add test for tls basic with mixed tls libs - testbench: fix python3 compatibility issues - testbench: fix parallel tests for receiver-abort.sh and other tests. - testbench: Added extra debug output to send.c utility. closes: https://github.com/rsyslog/librelp/issues/207 - testbench: obtain python binary path via AM_PATH_PYTHON closes: https://github.com/rsyslog/librelp/issues/214 - bugfix: librelp.h contains duplicate function definition - removed some more externally visible symbols not being part of API - testbench tooling improvements - distcheck bugfix: add missing testbench cert files closes: https://github.com/rsyslog/librelp/issues/206 ---------------------------------------------------------------------- Version 1.7.0 - 2020-08-25 - some internal cleanup (const attributes and such) - bugfix: library did export non-API symbols closes https://github.com/rsyslog/librelp/issues/187 - openssl: Fix chained certificate files for older OpenSSL Version. For older OpenSSL Versions, we are loading the Own Certfile (Which can be chained) into CTX Verify locations. This enables chained certificates for older OpenSSL Versions as well devtools: updated devcontainer.sh (based on rsyslog version) see also: https://github.com/rsyslog/librelp/issues/195 - fix FD leak when socket shutdown is one-sided AWS LB has timeouts when connection is inactive and when this happens, AWS LB shuts down socket from one-side only. When rsyslog detects one-sided shutdown, it assumes that connection is closed while it isn't. The connection is left unclosed and unclosed FD are piling up, eventually filling up FD number space. Thanks to Ognyan Kulev for the patch. - TLS: Added call to destruct OpenSSL remains to relpEngineDestruct For a clean cleanup, OpenSSL remains are now being removed at the end of relpEngineDestruct() closes: https://github.com/rsyslog/librelp/issues/199 - fix memory leak on session break When librelp session breaks unexpected, a memory leak could happen in sendq.c when relpSendqeConstruct was called before the session break happened. - Also Adds a new valgrind test basic-sessionbreak-vg.sh to reproduce the problem. without the fix, the test will fail. - Adjusted testbench plumbing to support session break valgrind test. closes: https://github.com/rsyslog/librelp/issues/188 ---------------------------------------------------------------------- Version 1.6.0 - 2020-04-21 - fix namespace pollution - some non-API functions were exported While fixing this, also some internal inconsistencies in function visibility were fixed as well. closes https://github.com/rsyslog/librelp/issues/179 - replsess: fix double free of sendbuf in some cases. In iRet handler of relpSessSendResponse, the sendbuf was freed if iRet returned a failure. However if error RELP_RET_IO_ERR happened in relpSendqAddBuf, sendbuf was already assigned to relpSendqe_t. As a result sendbuf was double freed in relpSendqDestruct. see also https://github.com/rsyslog/rsyslog/issues/4184 https://github.com/rsyslog/rsyslog/issues/4005 closes https://github.com/rsyslog/librelp/issues/183 - improve support for libressl LibreSSL has no support for @SECLEVEL and SSL_CONF_CTX_set_flags(). LibreSSL is now detected and some functionality disabled if it is used. Thanks to Remi Locherer for the patch. - Modified GnuTLS priority according to standard crypto-policy guideline Edited TLS priority string default to conform to system-wide default and align librelp behavior with other crypto-utilizing packages. Note: this may be different from what the openSSL driver does. Thanks to Jiri Vymazal for the patch. - tcp: Missing pUsr Copy to relpTcp Pointer fixed in relpTcpAcceptConnReq The userpointer was not copied into the relpTcp object. Also fixed core dump crash in receive / send helpers when userpointer is returned NULL from librelp. - report io errors for plain tcp connections this already happens with TLS, but not with plain tcp. Error messages are a useful troubleshooting tool. ---------------------------------------------------------------------- Version 1.5.0 - 2020-01-14 - bugfix: too late termination of relp Engine on shutdown When librelp is instructed to shutdown, it processes messages still present inside its receive buffers. It only terminates when it needs to wait for new data to arrive. Depending on RELP and TCP window size and message length, this may cause many messages to be processed while in shutdown. Even with default settings, up to 128 messages may be taken off the wire and be processed. This is a problem regarding the shutdown timing of a librelp user (e.g. rsyslog). It may take unexpectedly long to shutdown the RELP component and as such timeout may occur in the caller code. This is especially the case if the caller does lengthy processing when a RELP message is received. Note: It is perfectly fine for a caller to do this. The problem is that librelp continues to provide new data for a relatively long and unexpected period. This fix ensure that the relp engine shuts down much quicker when shutdown is requested. It now also checks the shutdown request while processing already received buffer data. This problem was detected when working on see also https://github.com/rsyslog/rsyslog/issues/3941 closes https://github.com/rsyslog/librelp/issues/175 - build system fix: invalid default in configure help text closes https://github.com/rsyslog/librelp/issues/169 - error message on invalid TLS library request added This way an invalid TLS library (name) can be detected and the error presented to the user. So far, invalid library names were hard to find. ---------------------------------------------------------------------- Version 1.4.0 - 2019-03-05 NOTE TO PACKAGERS: Both openssl and GnuTLS are now enabled by default. This is to enable apps to choice which of the two drivers they want to use. If you want to disable one or both of the drivers, use --disable-tls to disable GnuTLS (option name is historic) --disable-tls-openssl to disable openssl - build system: enable openssl by default, this means both TLS drivers are now build by default - support that both GnuTLS and openssl TLS drivers are active together This permits the caller to select which one to use via a NEW API relpEngineSetTLSLibByName() and relpEngineSetTLSLib(). - portability: use GCC __attribute__ only where supported - bugfix: build problem when HAVE_STRERROR_R is undefined Thanks to Paul Cameron for the fix. - bugfix: openssl driver did not properly handle retries when sending Every TLS library can request i/o call retry if in non-blocking mode. When sending, the openssl driver did not properly handle this. Note: the GnuTLS driver was not affected by this bug. - bugfix: in openssl mode, cert name validation did not work properly closes https://github.com/rsyslog/librelp/issues/149 - bugfix: invalid handling of connection fail could lead to abort ... or more precisely to SIGPIPE on the sender app. What exactly happens in that case depends on SIGPIPE handling in the sender. This is of course not acceptable or desired behaviour. This patch fixes it. This commit also includes testbench improvements, especially in regard to support easier diagnostic of such cases. closes https://github.com/rsyslog/librelp/issues/161 - tests: make dummyclient.py compatible with Python2 and Python3 closes: https://github.com/rsyslog/librelp/issues/163 Thanks to whissi for the patch. - a couple of minor and cosmetic nitfixes, improvements and cleanup ---------------------------------------------------------------------- Version 1.3.0 - 2018-12-11 - improved error reporting - bugfix openssl: anon mode did not work with openssl 1.1.0+ This was caused by "hardening" inside openssl, so not a real bug. - bugfix: do not send multiple open commands multiple open commands can happen when a session was broken and is being reestablished. Chances to experience this problem increase with the number of sessions broken in a row. A trigger is that the remote system accepts the TCP connection, but does NOT send a response to the "open" command (aka "very short-lived session). This can happen e.g. with proxy servers which cannot connect to the actual server. closes https://github.com/rsyslog/librelp/issues/129 - bugfix testbench: TLS tests were executed even when TLS was disabled - testbench: * add test for receiver abort and restart * much improved testbench framework and tools to support this type of test. Done some general improvements to the tooling. * imported rsyslog tool "chkseq" for checking ---------------------------------------------------------------------- Version 1.2.18 - 2018-09-18 - added non-standard "certvalid" auth mode to TLS authentication - bugfix CI: make distcheck did not work ---------------------------------------------------------------------- Version 1.2.17 - 2018-08-02 - added support for openssl now, at configure time, openssl (--enable-tls-openssl) or gnutls (--enable-tls) can be enabled. Openssl is recommended as it permits better error messages and has more robust certificate handling, but the default remains gnutls so that packagers do not need to change without specific need. Note: if openssl is selected, setting the gnutls priority string does not have any effect. This will be worked on in later releases. - improve code quality: replace strerror() by portable equivalent - improve error message on connection failure it now contains the OS-provided error message Thanks to Justin Azoff for the patch. - bugfix: 100% CPU utilization due to busy loop happend because EAGAIN was not properly handled Thanks to Lex Neva for the patch. closes https://github.com/rsyslog/librelp/issues/13 - bugfix: do not expose symbols that are not part of public API Some other symbols were exported. Now only symbols starting with "relp" are exported. Thanks to Michael Biebl for the patch. - bugfix: potential segfault when listener could not be bound This was fixed as a side-effect of refactoring some static analyzer warnings, exact commit is not flagged. see also https://github.com/rsyslog/rsyslog/issues/2874 ---------------------------------------------------------------------- Version 1.2.16 - 2018-05-14 - API changes * add new API: relpSrvSetOversizeMode() This permits to tell librelp how to handle oversize messages. Traditionally (and now default), this aborts the session. We now added an option to truncate the message instead. Also, in case of session abort a descriptive error message is emitted. This did not happen previously and caused confusion. closes https://github.com/rsyslog/librelp/issues/81 * add new API: relpSrvSetLstnAddr() It permits to set the listen address inside the relp server. If not called, the server will bind to all interfaces. Thanks to github user perlei for contributing it. - support additional hashes for fingerprint mode old-style SHA1 is used automatically Thanks to github user briaeros for the patch. see also https://github.com/rsyslog/librelp/pull/55 - bugfix: potential memory leak This is very unlikely to occur in practice. Memory can be leaked when TLS initialization fails when the client tries to connect to the server. However, if this actually happens, it can happen frequently and so accumulate to a large leak. No report of such occurence from practice. Detected by Coverity Scan, CID 266008. - bugfix: memory leak on protocol error Receiving relp frames are not correctly deallocated while handling protocol errors resulting in memory leak of dirty pages. Thanks to github user gleentea for the patch. see also https://github.com/rsyslog/librelp/issues/59 closes https://github.com/rsyslog/librelp/issues/60 - fixed a couple of minor issues: * fix memory leak when relp frame construction fails detected by clang static analyzer * removed unnecessary code detected by clang static analyzer * fix memory leak This leaks occurs if the process is already totally out of memory, a situation that is very rare and will also cuase other troubles. So the practical relevance of this patch looks rather slim. Detected by clang static analyzer. * fix memory leak on relpSrvRun() error this is kind of cosmetic, because it can only occur when the run fails, which usually should lead to termination of the calling application deteced by Coverity Scan, CID 266016 * fix memory leak on relp listener construction error detected by Coverity Scan, CID 266014, 266015 * also resolved all other issues reported by Coverity scan - CI * added native testbench (formerly used rsyslog for testing) * added additional compile tests ---------------------------------------------------------------------- Version 1.2.15 - 2018-03-21 - made build on Solaris again - made build on AIX Thanks to Philippe Duveau for providing the patches - bugfix: invalid handling of snprintf() return code - bugfix: invalid assert predicate an assert could change status variable due to typo, so in debug mode processing could fail. thanks to github user KatMisato for alerting us fixes https://github.com/rsyslog/librelp/issues/66 - some code cleanup - bugfix: error message on open error was truncated The "connection already open" error message when trying to open an already open connection was truncated due to too-small size specified. Thanks to rsyslog forum user AlanR for the problem report. ---------------------------------------------------------------------- Version 1.2.14 - 2017-05-29 - API extension: add relpSrvSetMaxDataSize() Thanks to Nathan Brown for the patch. ---------------------------------------------------------------------- Version 1.2.13 - 2017-02-20 - make build under BSD - bugfix: duplicated lines after server reconnect Thanks to Hervé for the patch. ---------------------------------------------------------------------- Version 1.2.12 - 2016-07-07 - API enhancement: permit to set connection timeout Thanks to Nathan Brown for the patch. - cleanup: replace deprecated GnuTLS data types by current ones - CI improvements: added dynamic tests we now use the rsyslog testbench to do dynamic librelp tests. This currently happens only under Travis CI. ---------------------------------------------------------------------- Version 1.2.11 - 2016-06-22 - bugfix: do not accept more than one "open" verb on a connection closes https://github.com/rsyslog/librelp/issues/37 - bugfix: potential segfault when high-numbered fd is used in sender When a fd (socket) with value >= 1024 was used by the client sender process, the library could segfault in select(). This depended a bit on the platform. This patch solves the issue by replacing the select() call with poll(). Note that we do not changed to epoll(), because (a) we only wait on a single fd (b) poll() is more portable closes https://github.com/rsyslog/librelp/issues/38 - bugfix: make librelp not run in thight loop when out of sockets If the process can no longer accept new connections (via accept) because it has reached its file handle maximum, librelp runs in a thight loop until either the client connection request is aborted or a file handle becomes available. This happens due to the fact that we get event notifications on connection requests. As we can't satisfy it, we librelp always retries. Unfortunately, there is no socket call to cancel a connection request, so we cannot push the client request out. An alternative solution would be to close and re-open the listening socket, but in that case it is uncertain if we ever can re-aquire a socket (an easy DoS). So the best thing we can probably do is delay retries a bit. We must be careful, however, not to use a too long delay, as that would also affect other connections. We could address that problem via a dedicated thread for the listening process, but librelp is designed so that it can be used by callers with any threading library and threading model, and that would violate this design principle. Judging the alternatives, a slight delay probably is the best solution, especially as the situation is uncommon and will hopefully not last too long. This patch implements that solution. closes https://github.com/rsyslog/librelp/issues/34 - bugfix: flush the current recv frame if it exists if the client session is restarting Thanks to Nate Brown for the patch. - API enhancement: add configurable connection timeout Thanks to Nate Brown for the patch. ---------------------------------------------------------------------- Version 1.2.10 - 2016-03-30 - make compatible with older versions of GnuTLS which require libgcrypt - improve example sender/receiver code Thanks to Chris Pilkington for the patches. - bugfix: Send full buffers after connection re-establishing When sending large buffers it's possible that only part of buffer data will be transferred before connection is closed. Then on connection re-establishing librelp thinks part of buffer is already sent and transfers only remaining part. Remote side then is not be able to parse such message and always closes the connection. Thanks to github user cellscape for the patch. ---------------------------------------------------------------------- Version 1.2.9 - 2015-12-15 - bugfix: Ignoring return status when handling syslog frames now. Otherwise valid messages in the frame buffer will get lost when the remote connection is closed during meantime. - build bugfix: GnuTLS header was required even if TLS was turned off Thanks to Pedro Alvarez Piedehierro for the patch. - bugfix: correct wrong assert() calls [debug mode only affected] ---------------------------------------------------------------------- Version 1.2.8 - 2014-09-07 - bugfix: segfault if KEEPALIVE is used Thanks to Tomas Heinrich for the patch. ---------------------------------------------------------------------- Version 1.2.7 - 2014-04-29 - bugfix: librelp was incompatible with C++ without a real reason Thanks to Ivan Kryak for the patch - bugfix: potential misadressing in wildcard match see: https://github.com/rsyslog/librelp/pull/3 Thanks to Sebastian Schmidt for the patch. - bugfix: always last wildcard match was reported, not first see: https://github.com/rsyslog/librelp/pull/4 Thanks to Sebastian Schmidt for the patch. ---------------------------------------------------------------------- Version 1.2.6 - 2014-03-20 - report error when preparing for non-anon TLS and this is unsupported The error is reported when the processes own certificate is set, which also turns off anon TLS. We now check if TLS AUTH is supported and, if not, return an error. This enables better error tracking in the client, as otherwise the problem is reported much later or even not at all (in the case of a relp server). ---------------------------------------------------------------------- Version 1.2.5 - 2014-03-20 - permit to use anonymous TLS on platforms where GnuTLS misses certificate verification function. This permits to use at least anon TLS on platforms like RHEL and CENTOS 6. ---------------------------------------------------------------------- Version 1.2.4 - 2014-03-17 - correct API/ABI change in 1.2.3 My reasoning was flawed, and we could run into problems with apps that required the new version but could not detect that an older one was installed. Thanks to Michael Biebl for pointing this out. What we have done is: - revert back to previous state (return void) * relpSrvEnableTLS(); * relpSrvEnableTLSZip(); These functions are now deprecated. - introduce new functions that return a state * relpSrvEnableTLS2(); * relpSrvEnableTLSZip2(); ---------------------------------------------------------------------- Version 1.2.3 - 2014-03-13 ************************************************************ * NOTE: this version should NOT be used due to the API * * change specified below. * ************************************************************ - add ability to build librelp without TLS This is required on some platforms where GnuTLS is too old Use --disable-tls to select this mode. If set, RELP_RET_ERR_NO_TLS will be returned on calls that selected TLS mode. See also: https://github.com/rsyslog/librelp/issues/1 - API change: two functions that used to return void now return state these are: * relpSrvEnableTLS(); * relpSrvEnableTLSZip(); This change is acceptable because the state must not necessary be evaluated. If not, the same error is returned a bit later in the calling sequence. Having it early, however, may help with better error messages. So it's still optional to check. - bugfix: configure required too-old version of GnuTLS This lead to successful builds which later on would not execute. ---------------------------------------------------------------------- Version 1.2.2 - 2014-01-07 - add capability to enable tcp KEEPALIVE - introduced new API relpSrvSetKeepAlive() to support KEEPALIVE ---------------------------------------------------------------------- Version 1.2.1 - 2013-09-26 - errors binding listener port are now reported via error message callback ---------------------------------------------------------------------- Version 1.2.0 - 2013-07-15 - support for epoll() added platforms that do not support it fall back to select() - API extension: relpEngineSetOnGenericErr ---------------------------------------------------------------------- Version 1.1.5 - 2013-07-05 - bugfix: memory leak on connection close around 60 bytes of memory were lost on each connection close at the server side (when the client initiated a close) - bugfix: potential misadressing on session close This can happen if also a write was outstanding, a quite unusable situation. In that case, already freed memory was being accessed. ---------------------------------------------------------------------- Version 1.1.4 - 2013-07-03 - fix build problems on BSD Thanks to Christiano for analysis and patch suggestion. - add ability to specify a non-standard RELP Window size - add burst support to the client API With this, we remove the unconditional performance improvement done in 1.1.3 - it is better to call the APIs explicitely. - wildcards are now supported in TLS name peer authentication They follow RFC5425 recommandations, e.g. *.example.com, client*.example.com, client01.*.com - new APIs: relpCltHintBurstBegin, relpCltHintBurstEnd, relpCltSetWindowSize ---------------------------------------------------------------------- Version 1.1.3 - 2013-06-26 - increased performance of RELP connection - bugfix: potential segfault if no GnuTLS priority string was set ---------------------------------------------------------------------- Version 1.1.2 - 2013-06-24 - add capability to specify the GnuTLS priority string This gives callers complete control over crypto parameters, like ciphers to use. - add certificate-based authentication - add capability to specify number of Diffie-Hellman bits to use - API extension: relpSrvSetDHBits, relpSrvSetGnuTLSPriString, relpSrvSetGnuTLSPriString, relpCltSetGnuTLSPriString relpEngineSetOnAuthErr, relpCltSetUsrPtr, relpSrvSetAuthMode, relpCltSetAuthMode ---------------------------------------------------------------------- Version 1.1.1 - 2013-06-11 - added compression support for TLS - API extension: relpCltEnableTLSZip, relpSrvEnableTLSZip ---------------------------------------------------------------------- Version 1.1.0 - 2013-06-03 - add TLS support - new api for creating listners: ... which permits us to set various properties before the listener is actually started. New callers should use it. Sequence is: * relpEngineListnerConstruct() * ... set properties ... (via relpSrv...() family) * relgEngineListnerConstructFinalize() This new style permits us to add/set additional listner options without the need to introduce ever-new listner create functions. Actually, their number would grow exponentionally, so this were a dead end. The old-style APIs relpEngineAddListner() and relpEngineAddListner2() are still fully supported (and supposed to be for a long time), but flagged as deprecated. ---------------------------------------------------------------------- Version 1.0.7 - 2013-05-13 - ABI change: removed relpCltConnect2() API which was against librelp API philosophy This was only introduced in 1.0.6 and been in the code for a very short time. So we decided that the best thing to do is actually remove it (there is NO known released user, this changes was for yet unreleased rsyslog 7.5.0). ---------------------------------------------------------------------- Version 1.0.6 - 2013-05-06 - enhanced API to permit binding a client to a specific IP address ---------------------------------------------------------------------- Version 1.0.5 (rgerhards), 2013-04-24 - bugfix: compile problem on Solaris Thanks to John Lyman for the patch ---------------------------------------------------------------------- Version 1.0.4 (rgerhards), 2013-04-11 - bugfix: busy loop on syslog sending as a client when server did not accept data; this was broken out only when the remote peer indicated that the connection as whole was broken. Now we properly timeout. ---------------------------------------------------------------------- Version 1.0.3 (rgerhards), 2013-04-10 - added relpCltSetTimeout() interface function - improved timeout handling Note: default timeout is now consistently 90 seconds (there were some parts of the code where it were 10 seconds and at others they were 180). - provide support for user-settable shutdown indicator This was introduced in support of rsyslog, but is probably useful for other users as well. In short, it provides the ability to set a pointer that is expected to be zero as long as the library shall continue to run and set to 1 if it shall terminate. That ptr is then checked at appropriate places of the library code. ---------------------------------------------------------------------- Version 1.0.2 (rgerhards), 2013-03-15 - added capability to support only IPv4/v6 instead of both ---------------------------------------------------------------------- Version 1.0.1 (rgerhards), 2012-09-06 - added capability to stop server without canceling its thread - bugfix: interrupt of select() was not properly handled However, this caused no actual failure, just some unnecessary code execution (and confusing debug messages). ---------------------------------------------------------------------- Version 1.0.0 (rgerhards), 2009-12-11 This version of librelp matured in practice and it is now time for a 1.0 release. Besides that, it includes a small number of changes: - bugfix: user callback never received remote IP address - bugfix: offers builder did use a fixed size string without bounds checking. I don't think this was a real issue as it was not exposed to the outside world, but now the buffer dynamically grows (which is the right thing to do). Thank to mterry for alerting me. ---------------------------------------------------------------------- Version 0.1.3 (rgerhards), 2009-02-10 - the callback on message reception did not contain a way to pass on a caller cookie (e.g. for an instance pointer). An additional interface has been added to support that. - cleaned up FDL license specifics - added libtool versioning - made librelp compile out of the box on FreeBSD - thanks to Michael Biebl for the patch ---------------------------------------------------------------------- Version 0.1.2 (rgerhards), 2008-04-09 - forward compatibility changes in support of our plans to use the rsyslog runtime in the future ---------------------------------------------------------------------- Version 0.1.1 (rgerhards), 2008-04-09 - improved build system, thanks to Michael Biebl now easier to integrate with autoconf ---------------------------------------------------------------------- Version 0.1.0 (rgerhards), 2008-04-01 - first version that is available to the public ---------------------------------------------------------------------- Version 0.0.1 (rgerhards), 2008-03-13 - very experimental and incomplete version, not to be used any longer