Forwarding to dbi-dev to make sure the patch doesn't get lost (it's
likely that the developers read dbi-users, but not certain).
This patch fixes a test failure with 15array.t on some platforms:
diff -Naur DBI-1.37/t/15array.t DBI-1.37-new/t/15array.t
--- DBI-1.37/t/15array.t 2003-05-11 13:45:27.000000000 +0100
+++ DBI-1.37-new/t/15array.t 2003-08-07 10:39:20.112963000 +0100
@@ -75,8 +75,11 @@
my $fetchrow = sub { # generate 5 rows of two integer values
return if $index >= 2;
$index +=1;
- # $index is quoted to avoid perl version differences
- return [ "$index", 'a','b','c' ];
+ # There doesn't seem any reliable way to force $index to be
+ # treated as a string (and so dumped as such). We just have to
+ # make the test case allow either 1 or '1'.
+ #
+ return [ $index, 'a','b','c' ];
};
@$rows = ();
ok( $sth->execute_array({
@@ -86,7 +89,8 @@
ok( @$rows, 2 );
ok( @$tuple_status, 2 );
$dumped = Dumper($rows);
-ok( $dumped, "[['1','a','b','c'],['2','a','b','c']]");
+$dumped =~ s/'(\d)'/$1/g;
+ok( $dumped, "[[1,'a','b','c'],[2,'a','b','c']]");
$dumped = Dumper($tuple_status);
ok( $dumped, "[1,1]");
Greg Earle pointed me to
<http://groups.google.com/groups?selm=20030530211041.GC27707%40dansat.data-plan.com&output=gplain>
which gives a possible fix for the test failure, but it seems that no
matter how hard you try to make $index a string, something along the
line (perhaps Data::Dumper) reserves the right to treat it as a
number. The only reasonable fix, it seems, is for the test suite to
accept both syntaxes.
--
Ed Avis <[EMAIL PROTECTED]>
