Compiling from Source --------------------- To compile the code the source or an installation of the Synthesis SyncML engine is needed. A compatible snapshot of it is included in SyncEvolution source packages and will be used automatically. The section _`Checking out the source` explains how to work with sources obtained via the git repositories. Also needed are the Evolution (can be disabled), Boost (>= 1.35) and pcre development files. For HTTP, either Curl or libsoup can be used. For compiling libsynthesis, sqlite, expat, libz are needed. On Debian based systems, the required packages can be installed with apt-get install evolution-data-server-dev \ libglib2.0-dev \ libecal1.2-dev libebook1.2-dev \ libsoup2.4-dev \ libpcre3-dev \ libexpat-dev \ libz-dev \ libboost-dev If gio (part of libglib) is not available in version 2.26 or higher, then libdbus can be used as fallback (not needed otherwise): apt-get install libdbus-glib-1-dev libboost-dev >= 1.34, available as libboost1.35-dev backport for Debian Etch. Necessary on some distros due to bad dependencies (not needed by SyncEvolution itself): apt-get install libdb3-dev Optional (enables reading proxy settings from GNOME preferences): apt-get install libsoup-gnome2.4-dev Optional (enables direct sync with phones): apt-get install libopenobex-dev libbluetooth-dev Optional (only used for SHA-256 when glib is not already a dependency): apt-get install libnss3-dev The test framework also requires CPPUnit: apt-get install libcppunit-dev For the GUI and its D-Bus based service backend: apt-get install xsltproc \ libglib2.0-dev \ libgtk2.0-dev libglade2-dev \ libgnome-keyring-dev \ libgconf2-dev libgnomevfs2-dev Optional packages for GUI: apt-get install libunique-dev libunique = ensure that GTK GUI only runs once per user Optional packages for GNOME Bluetooth Panel plugin: apt-get install libgnome-bluetooth-dev The plugin adds a button to invoke sync-UI after a device was paired which supports SyncML. The build system is the normal autotools system. See INSTALL for general instructions how to use that and "./configure --help" for SyncEvolution specific options. Note that compiling without the Evolution development files is possible. But because this is usually not what people want, the configure script needs explicit --disable-ecal --disable-ebook parameters, otherwise it will refuse to compile without Evolution support. When compiling from a git checkout, remember to run "./autogen.sh". It depends on: apt-get install libtool intltool automake Checking out the Source ----------------------- SyncEvolution is hosted on moblin.org. Anonymous access is via git clone git://git.moblin.org/syncevolution.git Before using sources checked out from Subversion, invoke "sh autogen.sh" with appropriate autotools packages installed. When running the configure script, it can be told to compile a Synthesis library automatically via the --with-synthesis-src configure option. This has advantages when modifying the Synthesis source together with SyncEvolution (no need to install Synthesis and thus faster compilation) and will bundle the Synthesis sources in source .tar.gz archives. The upstream Synthesis source code is here: git://www.synthesis.ch/libsynthesis.git The staging area for patches developed as part of Moblin are in the following repository: git://git.moblin.org/libsynthesis.git The intention is to include all these patches upstream to prevent forking the code. If you want to get patches included in the Synthesis code base, then please work directly with the upstream branch. For doing development work the recommended configure line is: configure --with-synthesis-src= \ --enable-warnings=fatal \ --enable-unit-tests \ --enable-libcurl \ --disable-shared \ --enable-developer-mode Enabling libcurl explicitly ensures that it gets built even when not the default. --disable-shared results in easier to debug executables (no shell wrapper scripts, all symbols available before the program runs). Backend libraries are dynamically scannned and loaded into syncevolution, the library path defaults to $prefix/syncevolution/backends. When developer-mode is enabled, it will scan libraries in current build directory instead. -Wno-unknown-pragmas is required to avoid warnings triggered by '#pragma }', a trick to preserve indention after 'extern "C" {' in /usr/include/evolution-data-server-1.12/libical/ Working with the Code --------------------- The code follows the code formatting of the Funambol C++ client library. Just emulate the existing code when possible. Exceptions derived from std::exception are used to report errors. The EvolutionSyncSource::handleException() function deals with logging the exception. SyncEvolution uses the a CPPUnit based testing framework. Configure with --enable-integration-tests and (optionally) --enable-unit-tests, then run "src/client-test" as described in src/client-test-app.cpp. It understands several environment variables, among them: - CLIENT_TEST_SERVER = chooses config - CLIENT_TEST_SOURCES = comma separated list of enabled sources, identified by their name For Evolution: eds_contact,eds_event,eds_task,eds_memo For file backend: file_contact,file_event,file_task,file_memo (uses same test data and vCard flavor as Evolution) - CLIENT_TEST_EVOLUTION_PREFIX=[name|file://] overrides the evolutionsource setting in the configuration; if file:// is used then these database will be created automatically NOTE : The automatic database creation is not supported for EDS 2.32. We need to create manually the databases using Evolution. New database names must be named ${CLIENT_TEST_EVOLUTION_PREFIX}${CLIENT_TEST_SOURCES}_[12]. Example: if CLIENT_TEST_EVOLUTION_PREFIX=Test_ and Client::Sync::eds_contact is used, then addressbooks with the names 'Test_eds_contact_1' and 'Test_eds_contact_2' must exist. - CLIENT_TEST_XML=1 use XML format as default instead of the normal WBXML For unattended testing: - CLIENT_TEST_FAILURES = comma separated list of tests which are allowed to fail without affecting the return code of the test runner; each list entry is a regex which must match a complete test name - CLIENT_TEST_SKIP = comma separated list of tests or test groups which are not to be executed at all; for this to work the test or test group has to be passed through test.h's version of ADD_TEST or FilterTest, which is the case for most tests but not all; as in CLIENT_TEST_FAILURES, each entry is a regex - CLIENT_TEST_LOG = name of server log file, will be copied and reset after each sync - CLIENT_TEST_ALARM = number of seconds a single test is allowed to run before aborting it Most Client::Sync tests use the default encoding, usually WBXML unless changed via CLIENT_TEST_XML=1. Client::Sync::*::testItemsXML always uses XML and Client::Sync::*::testItems always WBXML. Here are step-by-step instructions to get started with testing, using ScheduleWorld as example: - CLIENT_TEST_SERVER=scheduleworld \ CLIENT_TEST_EVOLUTION_PREFIX=file:///tmp/testing/ \ ./client-test -h => creates ~/.config/syncevolution/scheduleworld_[12]/ configs which use data bases under /tmp/testing, then prints all available tests - edit ~/.config/syncevolution/scheduleworld_[12]/spds/syncml/config.txt and enter account data for ScheduleWorld in both configurations; check that the syncURL is correct - CLIENT_TEST_SERVER=scheduleworld \ CLIENT_TEST_EVOLUTION_PREFIX=file:///tmp/testing/ \ ./client-test Client::Source => runs alls tests involving just local operations - CLIENT_TEST_SERVER=scheduleworld \ CLIENT_TEST_EVOLUTION_PREFIX=file:///tmp/testing/ \ ./client-test Client::Sync::vcard30::testCopy => runs one test that checks that one contact can be copied to and from the server using the two configurations - CLIENT_TEST_SERVER=scheduleworld \ CLIENT_TEST_EVOLUTION_PREFIX=file:///tmp/testing/ \ ./client-test Client::Sync => runs all tests which involve the SyncML server; tests involving just one source are run first, followed by the same tests with all enabled sources in two different orders "make valgrind" runs the same tests inside valgrind [http://www.valgrind.org]. A suppression file is used to hide errors inside system libraries which are not caused by SyncEvolution or Synthesis library code. Most likely the suppressions are system dependent and might have to be updated from time to time. Committing Changes ------------------ Here are some guidelines for well-formed commits: - Avoid unnecessary white space changes in the source code. - Don't mix unrelated changes in one single commit. Every single commit should leave the code in a functional state (compiles, unit tests pass, ...). - Document new functions and structures with Doxygen comments. These comments should really add information, not just repeat the name of the entity. At least paraphrase the names. - Each commit message should follow the format : [(BMC #1234)] - here refers to the module, class or functionality modified in the commit, for example "autotools" for the build scripts, "EvolutionContactSource" for the EDS contact backend, etc. - should be consise enough to describe the patch on its own. - The optional bugs.meego.com (BMC) number in round brackets refers to the issue addressed by the commit. Most changes, in particular bug fixes, should have such an issue filed for tracking purposes. - The body (hard-wrapped, multiple paragraphs for long text) should include the following pieces of information, if applicable: - problem statement (*why* is the change relevant?) - summary of the solution (*how* is the problem solved?) - what other solutions were considered and rejected (*context* for the solution) - known gaps (in particularly useful during code review, which will start with the oldest, possibly incomplete patch first) When writing source code comments and commit messages, always remember that you are communicating with someone. This person might be even be you in some distant future, trying to remember why the heck something was done the way it was done... There might not be a chance to ask questions, so think about what such a person might want to know and answer it in advance. There is a certain redundancy between source code comments, commit messages and issue tracker comments. Whenever possible, the description of the solution should be in the source code itself, with only a brief summary and/or pointer in the commit message. It's fine to copy-and-paste to reduce the overall effort. Building a Release ------------------ - increase version number in configure-pre.in/AM_INIT_AUTOMAKE For prereleases use "old version"+"new prerelease"-1 as package version. That ensures that package versions are higher than the old release, but lower than the final release. Avoid hyphens as part of the release names, i.e. use "0.7+0.8beta1" instead of "0.7+0.8-beta1". - ensure files were updated: ./NEWS debian/changelog - update po/LINGUAS - check and if necessary, bump the Synthesis .so versions (libsynthesis/src/Makefile.am.in) - ensure that Synthesis source code is tagged: use + if local changes exist - commit and tag SyncEvolution source code - make distcheck - compile binary .tar.gz packages for different Evolution versions; done automatically by runtests.py on estamos.de (= Debian 3.0), using different Garnome installations, and with special configure options to ensure maximum portability (LDFLAGS=-Wl,--as-needed --enable-static-cxx) - compile .deb for Maemo - update entries on the web about the release: http://maemo.org/downloads/product/OS2008/syncevolution/ http://www.estamos.de/blog/wp-admin http://www.estamos.de/projects/SyncEvolution/Roadmap.html http://freshmeat.net/projects/syncevolution/ Compiling for Maemo ------------------- unpack source archive in Scratchbox (for maximum compatibility: use Chinook 4.0 rootstrap; for support of all backends: ensure that the EDS-DBus calendar dev packages are installed), DEB_BUILD_OPTIONS=maemo fakeroot dpkg-buildpackage NOTE: dpkg-buildpackage -rfakeroot does *not* work as it leads to strange problems executing a.out during the client-src configure. Maemo EDS-DBus calendar dev packages /etc/apt/source.list: deb http://maemo.o-hand.com/packages chinook/ Compiling for iPhone -------------------- Requires iPhone toolchain and a libcurl compiled for the iPhone. libcurl is ideally configured as small as possible and statically (to avoid packaging problems): ./configure --prefix=/usr/local/iphone --host=arm-apple-darwin --disable-shared \ --disable-crypto-auth --without-gnutls --without-ssl --without-zlib \ --without-libssh2 --disable-ipv6 --disable-manual --disable-telnet \ --disable-tftp --disable-ldap --disable-file --disable-ftp manually set HAVE_POSIX_STRERROR_R in lib/config.h Potential problems with toolchain: std++ not found: ln -s libstdc++.6.dylib /usr/local/iphone-filesystem/usr/lib/libstdc++.dylib AddressBook framework must be added to iphone-dev/include/install-headers.sh.in Compile with curl-config in the PATH: PATH=/usr/local/iphone/bin/:$PATH ~/projects/SyncEvolution/trunk/configure --host=arm-apple-darwin --with-funambol-src=/home/patrick/projects/native CXXFLAGS=-O0 --disable-ecal --disable-ebook --enable-addressbook --prefix=/usr PATH=/usr/local/iphone/bin/:$PATH make all Build a package with: make distbin BINSUFFIX="iphone" Compiling for Mac OS X ---------------------- Configuring for development: /configure --with-funambol-src= \ --enable-addressbook \ --disable-ecal --disable-ebook \ CXXFLAGS="-Wall -Werror -Wno-unknown-pragmas" \ LDFLAGS="-framework Addressbook -framework CoreServices" \ CXXFLAGS=-g \ CFLAGS=-g Compiling final release: ./configure --enable-addressbook \ --disable-ecal --disable-ebook \ CXXFLAGS="-O -g -arch i386 -arch ppc" \ CFLAGS="-O -g -arch i386 -arch ppc" \ LDFLAGS="-arch i386 -arch ppc -framework Addressbook -framework CoreServices" \ --disable-dependency-tracking make BINSUFFIX=mac-os-x distbin Fine-grained memory checking: MallocStackLogging=1 MallocStackLoggingNoCompact=1 \ MallocScribble=1 MallocPreScribble=1 MallocGuardEdges=1 \ MallocCheckHeapStart=1 MallocCheckHeapEach=100 Debugging --------- The following packages contain debug information for relevant libraries on Ubuntu 7.10: evolution-data-server-dbg libglib2.0-0-dbg evolution-dbg Normally, syncevolution redirects stderr to its own log file. In crash scenarios the final error messages may get lost. To debug such cases, disable redirection by setting the environment variable SYNCEVOLUTION_DEBUG (value doesn't matter) and capture the output normally.