These were all caused by sprintf() appending a null char that was not actually used.
Fixes CVE-2021-30498 and CVE-2021-30499.

Fixes: #52 (CVE-2021-3410)

Fixes: #42

Fixes: #38 (CVE-2018-20546)
Fixes: #39 (CVE-2018-20547)

Fixes: #37 (CVE-2018-20545)
Fixes: #40 (CVE-2018-20548)
Fixes: #41 (CVE-2018-20549)

Fixes: #36 (CVE-2018-20544)

Hide private symbols (issue #33).

This will hopefully fix the OS X ruby package.

Fix a bug into the autorepeat trigger.

- Only affect slang/ncurses.

Issue #15: Fix returned pointer with caca types.

Fixed VS2015 error; snprintf std is now included

The Xlib and ncurses libraries query the LC_CTYPE locale value to guess
the usable character set. If the calling program did not call setlocale(),
that character set will be severely limited. Extensive Unicode support
is a reasonable libcaca user expectation.

The locale is restored as soon as possible, once the window or terminal
have been initialised. Unfortunately, the effect of setlocale() is process-
wide, and may affect other threads. This is now documented.

Note also that both Xlib and ncurses ignore the effects of uselocale()
which would have been the thread-safe solution to this problem.

This feature messes up with our Git checking history.

This fixes the path to source files in some unit tests and allows
to get rid of a few duplicate lines.

on the current ruby-dev instead of a specific version.

Fix python 3 issue

x11 driver: handle errors caused by missing locales