In order to use mouse to simulate touch input, we have to
generate multiple events on single SDL mouse click event.
It is every hard to achieve this in original implementaion
so I used pipe to fake a input dev as slider_watcher did.
Instead of linking against the location inside the original build tree
we should link against the copy in a single place. This makes the build
code cleaner and more compact even now and will do even more so when we
build more libraries as shared (which is what I am going to attempt
next).
after lots of cherry-pick I found that for some strange reason,
pic.c is not updated and I cannot find out which commit I missed.
So I simply copied the latest pic.c from the master branch here.
For now only line spaceing and word spaceing are configurable with
'F'/'Aa' commands. And device size are passed to k2pdfopt in place of
the hard-coded default width and height(via @dpavlin).
Conflicts:
koptreader.lua
1. In extr.c the error message should use the correct full pathname, not
the base filename.
2. In pdfattach the check for existence of input files should be
replaced with the check for read access to them.
1. Emit the list of attached files first and then emit each individual
attachment one per page.
2. Use hyperlinks in the list of attached files so one can jump directly
to the file which he wants to extract.
1. Set the page size to 9x12cm as appropriate for Kindle 3
2. Together with each attachment print its base filename and file size
in bytes.
3. Print the total size of all attachments (so one can estimate whether
it is safe to extract the attachments or not).
1. Handle filenames with spaces
2. Check if pdflatex and attachfile packages are installed before trying
to use them.
3. Use a more intuitive interface "pdfattach -o file.pdf file1.djvu
[file2.djvu]..." instead of forcing the user to specify "-i" option for
every input file.