tr: fix various issues with case conversion

This valid translation spec aborted:
  LC_ALL=en_US.iso-8859-1 tr '[:upper:]- ' '[:lower:]_'
This invalid translation spec aborted:
  LC_ALL=en_US.iso-8859-1 tr '[:upper:] '  '[:lower:]'
This was caused by commit 6efd1046, 05-01-2008,
"Avoid tr case-conversion failure in some locales"

This misaligned conversion spec was allowed:
  LC_ALL=C tr 'A-Y[:lower:]' 'a-z[:upper:]'
This was caused by commit af5d0c36, 21-10-2007,
"tr: do not reject an unmatched [:lower:] or [:upper:] in SET1"

This misaligned spec was allowed by extending the class:
  LC_ALL=C tr '[:upper:] ' '[:lower:]'

* src/tr.c (validate_case_classes): A new function to check
alignment of case conversion classes.  Also it adjusts the
length of the sets so that locales with different numbers of
upper and lower case characters, don't cause issues.
(string2_extend): Disallow extending the case conversion
class as in the above example.  That is locale dependent
and most likely not what the user wants.
(validate): Do the simple test for "restricted" char classes
earlier, so we don't redundantly do more expensive validation.
(main): Remove the case class validation, and simplify.
* tests/misc/tr-case-class: A new test to test the various
alignment and locale issues, associated with case conversion.
* tests/misc/tr: Move case conversion tests to new tr-case-class.
* tests/Makefile.am: Reference the new test.
* NEWS: Mention the fixes.

1 files changed, 0 insertions, 26 deletions

diff --git a/tests/misc/tr b/tests/misc/tr
index ca7a960bb..00cd8e62c 100755
--- a/tests/misc/tr
+++ b/tests/misc/tr
@@ -155,34 +155,8 @@ my @Tests =
 
   # Up to coreutils-6.9, this would provoke a failed assertion.
   ['no-abort-1', qw(-c a '[b*256]'), {IN=>'abc'}, {OUT=>'abb'}],
-
-  # Up to coreutils-6.9, tr rejected an unmatched [:lower:] or [:upper:] in SET1.
-  ['s1-lower', qw('[:lower:]' '[.*]'),
-   {IN=>'#$%123abcABC'}, {OUT=>'#$%123...ABC'}],
-  ['s1-upper', qw('[:upper:]' '[.*]'),
-   {IN=>'#$%123abcABC'}, {OUT=>'#$%123abc...'}],
-
-  # Up to coreutils-6.9.91, this would fail with the diagnostic:
-  # tr: misaligned [:upper:] and/or [:lower:] construct
-  # with LC_CTYPE=en_US.ISO-8859-1.
-  ['tolower-F', qw('[:upper:]' '[:lower:]'), {IN=>'A'}, {OUT=>'a'}],
-
-  # When doing a case-converting translation with something after the
-  # [:upper:] and [:lower:] elements, ensure that tr honors the following byte.
-  ['upcase-xtra', qw('[:lower:].' '[:upper:]x'), {IN=>'abc.'}, {OUT=>'ABCx'}],
-  ['dncase-xtra', qw('[:upper:].' '[:lower:]x'), {IN=>'ABC.'}, {OUT=>'abcx'}],
 );
 
-# Set LC_CTYPE=en_US.ISO-8859-1 in the environment of the tolower-F test.
-foreach my $t (@Tests)
-  {
-    if ($t->[0] eq 'tolower-F')
-      {
-        push @$t, {ENV=>'LC_CTYPE=en_US.ISO-8859-1'};
-        last;
-      }
-  }
-
 @Tests = triple_test \@Tests;
 
 # tr takes its input only from stdin, not from a file argument, so