Re: [PATCH 1/4] tests: exercise latent bug in character ranges

[Jim Meyering]

Paolo Bonzini wrote:
> * tests/bogus-wctob: New.
> * Makefile.am (TESTS): Add it.
...
ACK, modulo nits.

> diff --git a/tests/Makefile.am b/tests/Makefile.am
> index f354e4a..22bad3e 100644
> --- a/tests/Makefile.am
> +++ b/tests/Makefile.am
> @@ -39,6 +39,7 @@ TESTS =                                             \
>    backref-multibyte-slow                     \
>    backref-word                                       \
>    bre                                                \
> +  bogus-wctob                                        \

alphabetize?

>    case-fold-backref                          \
>    case-fold-backslash-w                              \
>    case-fold-char-class                               \
> diff --git a/tests/bogus-wctob b/tests/bogus-wctob
> new file mode 100644
> index 0000000..0237062
> --- /dev/null
> +++ b/tests/bogus-wctob
> @@ -0,0 +1,17 @@
> +#!/bin/sh
> +# This was latent in grep-2.8 and earlier because we punted interpretation
> +# of MBCSETs to glibc.  However, it becomes apparent as soon "[à]" is
> +# simplified to "à".

s/as soon/as soon as/

> +. "${srcdir=.}/init.sh"; path_prepend_ ../src
> +
> +printf '\340'      > in || framework_failure_
> +fail=0
> +
> +LC_ALL=en_US.UTF-8 grep '[à]' in
> +case $? in
> +  0) fail=1 ;;
> +  1) ;;
> +  2) framework_failure_ unexpected error from grep ;;
> +esac
> +
> +Exit $fail