Bug#835138: it's a matter of char signedness

[Adam Borowski]

Control: tags -1 +patch

Hi!
The failure comes from the assumption that char is signed.  That's true on
x86 and mips, but not on a number of other architectures.

A proper fix would require changing hundreds of places in the code; a quick
and dirty patch of compiling with -fsigned-char risks some trouble when
using external headers, but it appears to work for me -- at least it builds
ok and passes the testsuite.

Tested on arm64 and armhf, I assume the patch should also help on armel
ppc64el s390x powerpc ppc64; the FTBFS on x32 is unrelated.

-- 
⢀⣴⠾⠻⢶⣦⠀ Meow!
⣾⠁⢠⠒⠀⣿⡁
⢿⡄⠘⠷⠚⠋⠀ Collisions shmolisions, let's see them find a collision or second
⠈⠳⣄⠀⠀⠀⠀ preimage for double rot13!
-------------- next part --------------
A non-text attachment was scrubbed...
Name: form-uchar.patch
Type: text/x-diff
Size: 895 bytes
Desc: not available
URL: <http://lists.alioth.debian.org/pipermail/debian-science-maintainers/attachments/20170421/cd1f47bf/attachment.patch>