The problem is that before this change the same text could end wrapped differently
when written or when when it was pasted. When writing a text WebKit uses a
non-breaking space when pressing a space and then replacing it with regular space
if we continue with writing. Sometimes (after manipulating with selection) the
non-breaking space is not replaced with regular one thus the wrapping will fail
there. To avoid it we have to treat non-breaking space as a regular space.

If the anchor don't fit inside a line we have to always split it.

- If the position where we will split a line contains the dash character always
  split after it.
- Correct the check if we are already after the position limit.
- If ret_val is 0 also check if we really didn't found any character to break at
  before returning the limit position.
- Tight it to WebKitDOMCharacterData object.
- Rename the last_space variable as we don't split just on space.

Bug 753474 - Pressing Enter (Return) twice in a bulleted list to exit the list creates a paragraph with negative margin

Instead of leaving the operation on WebKit do it ourselves.

It adds an icon directory for one of the categories, which not always
works, especially when a translator in the evolution-data-server uses
a different category name translation, than the one from there. It is
better to traverse all categories and add all icon folders into the
search patch for the icon theme.

Avoid the crash by sticking to the GtkHTML based composer behavior. When there
are suggestions presented, automatically select the first one. If no suggestion
is presented, make the 'Replace' and 'Replace All' buttons insensitive.

The problem is that WebKit1 (2.4.9) crashes when it is trying to move or remove
an anchor element that has an image element inside and this image element has
the CSS float property set in the style attribute. To workaround it we will rename
the style attribute and rename it back when we will send the message. It is
unfortunate that we can change the formatting with this, but this is definitely
better than crashing. This could be removed once Evolution switches to WebKit2
as the WebKit2 is unaffected (tested on 2.8.4).

We don't support font size increments (+2,-2), but when going through a content
that was not written in Evolution we can find it. In this case just report the
normal size.

The regexec() match result returns byte offset, while gtk_text_buffer_get_iter_at_offset()
expects character offset, thus when the text contains any UTF-8 letters before the found URL
the link position was shifted by the UTF-8 bytes.

If we have an empty block (DIV element with just BR element inside) inside a
BLOCKQUOTE then the 'Blockquote' format is detected instead of 'Normal' as we
were getting a wrong node for inspection.

When opening a draft or a template with a signature we were setting the current
signature in "Signature chooser" to the one that was inside the message. The
problem was that the code replaced the body's signature with the one that is
saved on local storage. When the signature in the body was changed by the user
the changes were thrown away. After this change the signatures (the one in the
body and the one from local storage) are compared and if they are not the same
the signature in the body is left untouched and current signature in "Signature
chooser" is set to None (to stick with GtkHTML composer).

The code failed to get a next node to wrap when it previously processed a node
and its parent doesn't have a next sibling (this could happen in plain text mode
when we inserted a smiley).

If the user edited the signature or added more content after it, WebKit just
duplicated the DOM structure and left us with multiple PRE elements that don't
have the BR elements on their end. The content is rendered fine (every pre has
its own line), but when we try to get a plain text version of the signature we
will get the text from these PRE elements on one line. As a solution we need to
insert the BR elements on the end of earch PRE element (if not presented) to get
the correct text from signature.

We need to remove the 'data-user-wrapped' attribute from the block as otherwise
the BR elements that are used to wrap won't be removed in the
remove_wrapping_from_element function.