Re: [vile] strange 9.7 vile behavior under FreeBSD

[Thomas Dickey]

On Wed, 18 Jun 2008, Gary Jennejohn wrote:

> Hello,
>
> I've updated the FreeBSD vile port to 9.7 and now I'm observing _very_
> strange behavior.
>
> If I don't make vile with ICONV support, then when I start it, it
> totally ignores all keyboard input.  Note that it comes up just fine and
> has also loaded the (in my test shmode) appropriate filter.
>
> The last version I'd updated the port to was 9.6e and I know that vile
> worked just fine without ICONV support there.  The 9.7 xvile also works
> without ICONV support turned on.
>
> I've found a workaround - if ICONV is turned off then if I also turn
> off i18n support (--without-iconv --without-locale) then vile works OK.
>
> Conversely, if ICONV is turned on then I also turn on i18n, but I'm not so
> certain that's the correct thing to do.  But --with-locale seems to be the
> default anyway. which seems questionable to me.
>
> This all seems kind of strange though.
>
> Comments?
>
> BTW my locale settings:
> garyj:peedub:~:-bash:35> locale
> LANG=
> LC_CTYPE="C"
> LC_COLLATE="C"
> LC_TIME="C"
> LC_NUMERIC="C"
> LC_MONETARY="C"
> LC_MESSAGES="C"
> LC_ALL=

Revisiting this - as exactly as I can with my 6.0 FreeBSD, I'm still
not seeing a problem.  I set the locale the same, and am indeed linking
to iconv (different versions of course):

../bin/vile:
        libiconv.so.3 => /usr/local/lib/libiconv.so.3 (0x28122000)
        libncurses.so.6 => /lib/libncurses.so.6 (0x2820f000)
        libperl.so => /usr/local/lib/perl5/5.8.7/mach/CORE/libperl.so (0x2824e0>
        libm.so.4 => /lib/libm.so.4 (0x28350000)
        libcrypt.so.3 => /lib/libcrypt.so.3 (0x28366000)
        libutil.so.5 => /lib/libutil.so.5 (0x2837e000)
        libpthread.so.2 => /usr/lib/libpthread.so.2 (0x2838a000)
        libc.so.6 => /lib/libc.so.6 (0x283af000)

I didn't see anything useful in my log-comments for eightbit.c, but
as I mentioned, I seem to recall fixing during development some issue
that would cause vile to ignore input.  That's a thousand lines of
code though (I don't recall whether it was sign-extension or checking
return-values from iconv).

--
Thomas E. Dickey
http://invisible-island.net
ftp://invisible-island.net