Just a bit of the beginning of "tuxpaint.1" has been
wrapped in localizing "gettext()" calls. Generating
snapshot of new English version, as well as Spanish,
French, Galician, and Japanese, based on the PO files
that exist over in `tuxpaint-docs` repo.
Ensured Tux Paint's built-in help ("tuxpaint --help"),
Unix manual ("man page"), and command-line option
tab-completion list all cover the full set of Tux Paint
options. Some were missing in a few places.
Luc Schrijvers reported on the Tux Paint Maintainers mailing list an
failure to build magic tools for Tux Paint 0.9.26 rc1 (see the mailing
list for the details). The issue appears to stem from an interesting
interaction between my commit from March
(39cc096ece) and the quotes around
beos_PLUGIN_LIBS on line 1361:
beos_PLUGIN_LIBS:="$(MAGIC_SDL_LIBS) $(MAGIC_ARCH_LINKS) $(MAGIC_SDL_CPPFLAGS)"
My commit had removed a leading argument from $(MAGIC_SDL_LIBS) (whose
value is irrelevant but it was "-L/usr/local/lib"), and the second
argument resolves to nothing on Haiku, so beos_PLUGIN_LIBS now resolves
to a string with a leading space, and it appears `cc` treats a string
argument with a leading space as a single token as opposed to a list of
arguments as it was apparently doing previously. Pere confirms removing
the second argument that resolves to nothing allows binary to compile.
I suppose removing the quotes around beos_PLUGIN_LIBS is another (and
cleaner) possible solution, but the quotes appear to be intentional and
without a Haiku build environment to test the exact behavior this will
be the safer commit to make.
simulate_flood_fill_outside_check() could spin in an
infinite loop because it could sometimes set values in the
`touched[]` array to 0, which means the spot was never seen.
Values going in there are not boolean, but a 0-255 gradient,
for smoother fills on antialiased edges.
This change forces it to be a 1-255 value, ensuring that
a 'touched' spot on the canvas will have a non-zero `touched[]`
value.
h/t Pere for noticing the bug, providing some details,
and best of all providing a sample image to play with which
consistently invoked the crash. (It made mending this a
10 minute task for me, rather than an hours-long hunt to reproduce!
Big 'thanks' for doing that ahead of time!!!)
