GtkBuilder no longer sets widget name based on ui definition id. This
breaks name-based styling on all widgets originating from Glade, as
Glade does not take the GtkBuilder change into account.
Work around this by explixitly setting the widget names for the widgets
that are referenced by name in the rc-file.
The Sony Ericsson W595 only support contacts and events, but no tasks.
That triggered an access to a "NoneType" for the task config. Need to
check not only for existance of key, but also for real value.
client-test now accepts test case files like ical20.ics.<server>,
without the .tem suffix. This is used for testing the problematic
VCALENDAR items from MBC #1332, like this:
CLIENT_TEST_SERVER=local ./client-test Client::Sync::ical20::testConversion
Because "local" is not a recognized server, one has to set the syncURL
in the created "local_1" config before the test runs.
With the server properly recording per-source information, it is
possible to check for it. In the "client disconnect" case, the
overall status is "transport failure" whereas the one source which
was activated ends up with "aborted on behalf of user".
The latest Synthesis engine also emits progress events in server mode.
Before this patch, the events were ignored because the code was
conservative about not relying on something not supported. Now
client and server mode report and record progress and statistics
the same way.
The main reason for bundling the .html version with the distribution
packages is that the syncevolution.org "Usage" page needs the
.html version. Building it as part of of the distribution packages
ensures that packages and documentation are in sync.
An additional benefit is that some users might prefer to read the
documentation in HTML.
As with man pages, building the HTML README depends on rst2html and
can be controlled via --with-rst2html.
This patch adds the necessary autotools rules a) to use README.rst
as README in distribution archives and b) to build a syncevolution.1
man page out of it.
The later depends rst2man. --with-rst2man can be used to force man page
creation on or off. Without an explicit choice, it is only built when
the tool is found.
Version and compile date are inserted automatically.
Reorganized the content so that it is more like a man reference page.
Known problems, setup, feature overview are meant to be maintained
at syncevolution.org.
Changed the format to reStructuredText (http://docutils.sourceforge.net/rst.html).
It can already be converted to man format (rst2man) and HTML for the
Drupal installation on syncevolution.org. Line breaks are problematic
when copying the HTML source because Drupal turns line breaks into <br/>.
The goal is that the README.rst is the authoritative copy of the
syncevolution.org "Usage" page and a "syncevolution/synccompare" man
page.
After switching to files for template, temporary files as
created while editing a template file were also read as templates,
leading to duplicate templates (one old, one modified).
With this patch, invisible files starting with a dot and backup
files ending in a tilde are ignored.
With this patch, template files are also found in
.config/syncevolution-templates (or the corresponding directory
pointed to by the XDG env variables).
The tests had to be adapted a) to not find templates installed
by the user in his own home and b) to cover finding templates
via XDG_CONFIG_HOME.
This patch switches reading of templates on disk from the previous,
one file per .ini file format to the single-file format produced
by the syncevo-phone-config utility. This makes creating templates
much easier, also when using just an editor instead of the tool.
The existing Cmdline tests cover this code pretty well, as demonstrated
by the number of fixes that were necessary in the previous patches
when running the tests. Therefore it should be safe to use this in
SyncEvolution 1.0.
This patch adds the infrastructure for reading a single file
which contains multiple separate .ini files. Writing is not
implemented.
The Ini*ConfigNode classes are derived from the File*ConfigNode
classes and could replace those. The only reason for not doing
this right away is that the master branch is in code freeze in
preparation for 1.0. Removing File*ConfigNode and replacing
their usages should be done after the release.
The SingleFileConfigTree implements the splitting of the single
file into independent pieces. It instantiates IniFileConfigNodes
with in-memory access to the actual data, using the new classes
introduced for this case in the previous patches.
SingleFileConfigTree could also use other ConfigNodes, but that kind
of flexibility is not needed yet and also would require rethinking the
way how the single file is split. Right now the separators ("===
.... ===") are known to not occur in the individual pieces.
Like the FileConfigTree, this class must be able to preserve
nodes which were created without flushing the tree afterwards.
This is done when reading from a read-only template, which may
add new nodes or modify existing ones in memory.
This class works with a string whose ownership is shared with
others. The goal is that this will avoid copying strings around.
Right now, writing into that string is not implemented, because
only reading from an existing string is needed for the single
.ini file use case (MBC #1208).
The goal of the DataBlob base class is encapsulating read/write
access to permanent data storage *on demand*. Comparing it to
files, it is more like a path name than a file handle.
boost::shared_ptr are used a) because it automates deletion of
the actual data streams and b) some implementations of DataBlob
might have to share access to these instances.
One possible implementation (included in this patch) is indeed
a plain file. It writes into a temporary file and only renames
it to the final path name on success, which emulates the behavior
of the FileConfigNode. The actual implementation of this renaming
was moved into a separate class, the SafeOstream.
Another (to be added later) is a chunk of data inside a larger file.
This is needed for embedding different .ini style files inside
one large text file, as required for MBC #1208).
Added functions to split a full path name into its components.
The base name is implemented as in the "basename" utility,
i.e., the trailing slash in "foo/" is ignored, leading to "foo"
as base name.
Types like "text/calendar" are implicitly turned into
"text/x-vcalendar" inside SyncEvolution when asking the phone to
sync. Forcing the type leads to a different SAN message and thus
is worth testing.
Forcing legacy types also leads to a slightly different sync session:
the SAN is the same as without forcing the type, but then we send
simpler CtCap information. This might make a difference when a phone
stumbles over too complex CtCap. But because such a phone should
work when disabling CtCap altogether, the forced legacy types are
not tested separately (would increase overall number of test cases).
Using --advanced with a --type for which no test data exists now fails
with an error instead of silently using empty items.
The Sony Ericsson K750i accepts "text/x-vcalendar" as type in the SAN message,
but not "text/calendar". We had code which was supposed to map to the older
type, but due to a typo ended up sending "text/x-calendar", which is rejected.
Fixed.
The code which set remoteIdentifier and SyncMLVersion in the template
copied from a "config" variable which was not set if a) not creating
a config earlier or b) the last source had no working config.
Case b actually occurred when running against a MeeGo netbook, which
has no "memo" database.
The fix is simple: only copy the values when at least one source has
them.
