This fixes a segfault in --reload when the image becomes unloadable while feh
is running.

Note that this does not affect --reload with more than one file: If the
current image is removed, feh will simply jump to the next image. feh will
only quit if the file still exists, but can no longer be loaded.

This fixes half of the issue - windows on Xinerama screen 0 no longer become
too large, but the bug still appears on the other screens (unless
XINERAMA_SCREEN is set)

* remove unused variable Im1 in filelist.c
* options are initialised as 0, so only explicitly set those to 0 for which it
  makes sense because of their context
* Make a few warnings about wrong option combinations fatal

Note that read_filelist used to check if its argument was an image (for people
accidentaly typing "feh -df *" instead of "feh -dF *"); this was removed
because it failed on /dev/stdin.

I might re-add it, but I'm not yet sure about the future of feh's automatic
filelist saving.

feh_draw_info used to iterate line-wise over the info output, but now we need
to cache it somewhere to get the length of the longest line.

67e41d60 moved the ZOOM_MODE from options.h
defines to a feh.h enum, but did not set the values correctly.

It is pretty random and I doubt anyone actually uses it. Also, after removing
it, there are only two duplicates of the image index code instead of three...

The info text used to be located 2px away from the bottom / left edges.  Now
that --draw-tinted has been implemented, that leaves us with an ugly edge, so
move it directly to the edge.

By default, feh draws white text with a black drop-shadow, which is well
readable in most, but not all cases.  The --draw-tinted option draws a
semi-transparent black box around all texts.  This improves readability, but
does not look as nice, so it won't become default behaviour.

It was suggested to use #7f7f7f7f for the background, but I decided that
for small images surrounded by a black border.

feh_draw_info used to unconditionally swallow the last character of a line,
assuming that it must be a newline. However, the very last line of a program's
output may not end in a newline, in which case its last output character was
omitted.

An action may be used to alter the image or specific parts of it displayed by
--info, so if a user invokes an action with hold_action set, the image and all
relevant information must be reloaded.

In the image_remove function, slideshow_change_image needs to be called before
removing the image from the filelist.  Because of this, the "x of y" displayed
by --draw-filename is wrong (y is 1 too high).

This commit introduces a new 'render' argument to slideshow_chingae_image.
The image_remove function calls it with render=0, edits the filelist and then
runs by winwidget_render_image itself.

Patch by Yu-Jie Lin.

gib_imlib_save_image_with_error_return was used, but it tries to set the Imlib
image format by examining the filename to save to.  However, it only supports
lowercase file endings (as in .jpg) and breaks with others (like .JPG).

Since giblib is unmaintained, for now the fix is only used here.

Signed-off-by: Yu-Jie Lin <livibetter@gmail.com>

A page is the window width or height.

Signed-off-by: Yu-Jie Lin <livibetter@gmail.com>

This fixes a really weird bug where continued theme definitions containing
just one option/value pair caused feh to mis-parse parts of the following
line.

Apparently, subsequent realloc calls mixed with strdups in another function
caused parts of the list (argv) content to be overwritten by the content of
the list pointer itself. I wasn't able to find out the exact causes /
conditions.