[perl.git] branch blead, updated. v5.23.5-248-g157a0ac

[Jarkko Hietaniemi]

In perl.git, the branch blead has been updated

<http://perl5.git.perl.org/perl.git/commitdiff/157a0ac7a2e2f56bd291b87c343712def4a1acfb?hp=709be747a32edc503b4645d9c5396bd4b40100d2>

- Log -----------------------------------------------------------------
commit 157a0ac7a2e2f56bd291b87c343712def4a1acfb
Author: Jarkko Hietaniemi <j...@iki.fi>
Date:   Wed Dec 16 21:57:31 2015 -0500

    Configure: notes on the m68881 extended precision format

M       Configure

commit 1c27387306f6b9c92ba951acd5641e71039e8352
Author: Jarkko Hietaniemi <j...@iki.fi>
Date:   Fri Dec 18 07:19:12 2015 -0500

    Double-double implementations differ.

M       t/op/sprintf2.t
-----------------------------------------------------------------------

Summary of changes:
 Configure       | 10 +++++++++-
 t/op/sprintf2.t |  8 ++++----
 2 files changed, 13 insertions(+), 5 deletions(-)

diff --git a/Configure b/Configure
index 2a4ee8c..0e71b4b 100755
--- a/Configure
+++ b/Configure
@@ -6967,7 +6967,15 @@ int main() {
     exit(0);
   }
   if (b[0] == 0xBF && b[9] == 0xCD) {
-    /* is there ever big-endian 80-bit, really? */
+    /* Is there ever big-endian 80-bit, really?
+     *
+     * The Motorola 68881 had another "extended precision" format:
+     * sign:1 exp:15 zero:16 integer:1 mantissa:63
+     * for total of 96 bits of bytes.  The zero bits were unused.
+     * See "M68000 FAMILY PROGRAMMER’S REFERENCE MANUAL" for more details.
+     * If it ever becomes relevant, this format should be allocated
+     * a new doublekind code since it's quite different from the Intel x87.
+     */
     printf("4\n");
     exit(0);
   }
diff --git a/t/op/sprintf2.t b/t/op/sprintf2.t
index be16c37..2784bde 100644
--- a/t/op/sprintf2.t
+++ b/t/op/sprintf2.t
@@ -674,10 +674,10 @@ SKIP: {
         unless ($Config{uselongdouble} &&
                 ($Config{longdblkind} == 5 ||
                  $Config{longdblkind} == 6)
-                # TODO: gating on 'linux' here is only due to lack of
-                # testing in other big-endian platforms (e.g. AIX or IRIX),
-                # with more evidence this subtest could be either relaxed
-                # or removed.
+                # Gating on 'linux' (ppc) here is due to the differing
+                # double-double implementations: other (also big-endian)
+                # double-double platforms (e.g. AIX on ppc or IRIX on mips)
+                # do not behave similarly.
                 && $^O eq 'linux'
                 );
     # [rt.perl.org 125633]

--
Perl5 Master Repository