Apparently the migration logic was more complex than it needed to be.
The old numeric key was already synced to the EMailReplyStyle enum in
the source code. Dunno where I got the idea it wasn't.
Just more evidence numeric enum keys are bad.
Take a CamelStore and folder name instead of a CamelFolder.
CamelStore and folder name can easily be obtained from either a folder
URI or a CamelFolder instance, and the function is more efficient with
separate parameters.
We were using g_object_get() to write an "unsigned int" value (at least
32 bits) into a 16-bit integer address.
Don't know why we were bothering with g_object_get() in the first place,
just call camel_network_settings_get_port() instead.
My previous patch didn't go far enough to repair an already-corrupted
"show-headers" key. Need to actually reset it to its default value if
the "headers" key is empty.
SpamAssassin is clearly not intended for use by mail clients, as
evidenced by the number of backflips Evolution's SA module still has to
do just to detect the presence and nature of a running spamd (a problem
which D-Bus solved a decade ago), and recent SA developer comments.
In lieu of removing SA support entirely, remove all the crazy GSettings
that are (thankfully) not exposed in the UI and relegate Evolution's SA
integration to only the most basic usage (spamassassin / sa-learn).
Users are better off with Bogofilter anyway. Leave SpamAssassin for
mail servers.
EShellView no longer needs help from subclasses other than getting
the needed GalView subclasses registered.
A nice side-effect of this is EShellView subclasses can now use the
G_DEFINE_DYNAMIC_TYPE macro.
No longer needed. Instead, use g_type_ensure() to ensure the necessary
GalView subclasses are registered in the GType system before loading a
GalViewCollection. Best place to ensure types is from GClassInitFunc.
EShellView now holds a reference to the active GalViewInstance. Where
applicable, the EShellView subclass is responsible for keeping this up
to date when the sidebar selection changes.
Holding a reference allows EShellView to implement common actions like
"Save Current View" directly instead pushing it on to subclasses.
New functions:
e_shell_view_get_view_instance
e_shell_view_set_view_instance
e_table_specification_new() now takes a table specification filename and
a GError and parses the file as part of instance creation. If a file or
parse error occurs, e_table_specification_new() returns NULL.
This replaces e_table_specification_load_from_file().
New functions:
e_table_specification_get_filename()
Removed functions:
e_table_specification_load_from_file()
I don't remember why E_TASK_TABLE_DEFAULT_STATE was needed when
e-calendar-table.etspec already specifies a default ETableState.
Some of the column numbers it was referencing weren't even valid.
Removing it seems to make no difference in Evolution.
I don't remember why E_MEMO_TABLE_DEFAULT_STATE was needed when
e-memo-table.etspec already specifies a default ETableState.
Removing it seems to make no difference in Evolution.
The shortcuts Ctrl+C/V/X are used for whole calendar items
copy/paste/cut, not for text when editing event details inline, either
in a day/week view or in a list view. By tracking the is-editing property
of respective cell editor and using it when enabling/disabling clipboard
actions makes the respective text operations work as expected.
This commit does a number of things which I could not subdivide into
smaller commits.
* Converts ETreeModel to an interface, implemented by MessageList.
* Drops ETreeMemory and ETreeMemoryCallbacks, which were ETreeModel
subclasses. Their functionality is subsumed by MessageList.
* MessageList drops its public ETreeModel pointer, since MessageList
now implements ETreeModel as an interface.
* Adds message_list_set_expanded_default(), which takes over for
e_tree_memory_set_expanded_default().
The returned UID array now has a built-in "free" function for its
elements and should be released by callers with g_ptr_array_unref()
rather than em_utils_uids_free() or some equivalent.
Also add a placeholder ESettingsMessageList extension. Going to clean
out some of the direct GSettings usage in MessageList by adding GObject
properties and binding them to GSettings keys from the extension.
This is another source of crashes after creating and destroying a second
shell window. The signal handler was left connected with the destroyed
shell window as the closure.
This fixes a bug in the old implementation where the application could
crash after a second shell window was created and destroyed, because a
signal handler with the destroyed shell window as the closure was left
connected.
But moreover this simplifies the implementation by using a property
binding plus transform function instead of juggling signal handlers,
and also adds code comments where things get a little tricky.
Removed (now unused) functions:
e_shell_window_get_toolbar_new_prefer_item
e_shell_window_set_toolbar_new_prefer_item
Call e_source_selector_update_all_rows() after connecting to the
selector tree model's "row-changed" signal. This will ensure the
appropriate ECalClients get loaded into the ECalModel.
This fixes the bug where opening a second Tasks window shows no
content until one of the sidebar items is fiddled with.
Call e_source_selector_update_all_rows() after connecting to the
selector tree model's "row-changed" signal. This will ensure the
appropriate ECalClients get loaded into the ECalModel.
This fixes the bug where opening a second Memos window shows no
content until one of the sidebar items is fiddled with.
Call e_source_selector_update_all_rows() after connecting to the
selector tree model's "row-changed" signal. This will ensure the
appropriate ECalClients get loaded into the ECalModel.
This fixes the bug where opening a second Calendar window shows no
content until one of the sidebar items is fiddled with.
For messages with a base MIME type of multipart/alternative, we were
hiding text/plain subparts based on the number of alternate subparts.
This assumption of course broke on a message with the following body
structure and a Plain Text Mode preference of "Show HTML if present":
multipart/alternative
text/plain
text/plain
Instead, note when we've actually seen a text/html subpart and use that
to decide whether to hide the text/plain parts.
This will replace the headers API in EMailFormatter. Need a more
permanent place for headers since EMailFormatter is too disposable.
Also add an ESettingsMailPartHeaders class, which binds the new property
to the "show-headers" setting with a suitable mapping function to filter
out disabled header names.
Replaces the "headers" key. Whereas "headers" is an array of XML
blobs, the "show-headers" key is an array of string/boolean pairs.
Also update the appropriate places to get/set the new key.
Mainly to avoid accessing GSettings directly from EMailBrowser.
Also add a "browser-close-on-reply-policy" GSettings key that replaces
"prompt-on-reply-close-browser", the difference being the new key uses
an enum definition compatible with EAutomaticActionPolicy instead of a
free-form string value.
And finally add an ESettingsMailBrowser class to glue things together.
