1. When loading boolean settings we have to explicitly check for the
return of nil from the self.settings:readSetting() function and only
modify the default if the return value is NOT nil.
2. When closing the document we have to reset the settings to defaults,
otherwise their values are re-used on the next open.
3. The messages "Links on page ON/OFF" changed to
"Link underlines ON/OFF" because the old version gave two false
impressions: a) as if the action is pertinent only to the current page
(but in fact it is global for the entire document) and b) that we are
disabling the links whereas we are only disabling the show of underlines
under the links. The new message addresses both issues.
The function FileChooser:FullFileName() was calculating the position of
the item incorrectly, i.e. would work only on the first page in
filechooser. This would break copying/moving files to clipboard and also
renaming files from the second page and subsequent ones.
This is to make adding support for new formats easier.
See the file pic_jpeg.h for the description of the interface that each
image format must support.
If the number of colour components is neither 1 nor 3 we should remember
to free the buffer holding the raw decoded image before returning the
error to Lua.
1. Rewritten the function FileChooser:setPath() to handle returning to
the parent directory via ".." entry correctly. Issue #419
2. Fixed the crash when moving the last file from a directory to the
clipboard via Shift-X and then pressing Shift-X on the phantom entry
that is left. Issue #420
On some distributions (e.g. Ubuntu) the shell used for scripts is dash,
not bash and we explicitly rely on bash's features in the Makefile, so
we must guarantee that we are running with bash (not dash or anything
else).
the checking of the boundaries when blitting was incomplete
for the addblitToBuffer() function when compared with the
blitToBuffer() function. Factored out the redundant checking
procedure into a new function.
Should fix another bug leading to issue #401 and probably
more cases for specifying boundary boxes.
My fixes to kerning logic check in sizeUtf8Text(), renderUtf8Text() and
renderUtf8TextWidth() functions either caused or triggered a memory
corruption easily observable with DjVu files.
Undoing these changes fixes the corruption... or does it just make it
unnoticeable? Please review these changes to find out the answer.
