Support gcc-4.x on HP-UX PA-RISC/64

Correct the socketsize. Probably due to big-endian versus
little-endian, this has always worked on Intel CPUs. This
is a very very old problem,  and it has been the reason I
never used gcc-4 on PA-RISC,  because perl would not pass
the test suite in 64bitall.

Noticeable effects of wrong "$socksizetype" are return codes
of 0 (pass) from functions like getpeername (),  but invalid
(or none) data in the returned structures or a return length
of 0 or 256. In the latter case, the length is stored in the
second part of the 64bit long and the 32bit first part is 0.
This might be true on Intel-like machines too, but the tests
will pass, as the significant part of the returned length is
in the first 32bits and a pointer to int will still see that
correct.  Mind that in that case the 32bits after that might
have been overridden => can of worms.

diff --git a/ext/Socket/t/getaddrinfo.t b/ext/Socket/t/getaddrinfo.t
index 5c8d153..b85af38 100644 (file)
--- a/ext/Socket/t/getaddrinfo.t
+++ b/ext/Socket/t/getaddrinfo.t
@@ -1,6 +1,6 @@
 use strict;
 use warnings;
-use Test::More tests => 29;
+use Test::More tests => 30;
 
 use Socket qw(
    AI_NUMERICHOST AF_INET SOCK_STREAM IPPROTO_TCP
@@ -21,9 +21,15 @@ is( $res[0]->{socktype}, SOCK_STREAM,
    '$res[0] socktype is SOCK_STREAM' );
 ok( $res[0]->{protocol} == 0 || $res[0]->{protocol} == IPPROTO_TCP,
    '$res[0] protocol is 0 or IPPROTO_TCP' );
-is_deeply( [ unpack_sockaddr_in $res[0]->{addr} ],
-           [ 80, inet_aton( "127.0.0.1" ) ],
-           '$res[0] addr is {"127.0.0.1", 80}' );
+ok( defined $res[0]->{addr},
+   '$res[0] addr is defined' );
+if (length $res[0]->{addr}) {
+    is_deeply( [ unpack_sockaddr_in $res[0]->{addr} ],
+              [ 80, inet_aton( "127.0.0.1" ) ],
+              '$res[0] addr is {"127.0.0.1", 80}' );
+} else {
+    fail( '$res[0] addr is empty: check $socksizetype' );
+}
 
 # Check actual IV integers work just as well as PV strings
 ( $err, @res ) = getaddrinfo( "127.0.0.1", 80, { socktype => SOCK_STREAM } );

diff --git a/hints/hpux.sh b/hints/hpux.sh
index cb17184..be6c1fd 100644 (file)
--- a/hints/hpux.sh
+++ b/hints/hpux.sh
@@ -263,6 +263,16 @@ EOM
 
        case "$ccisgcc" in
            $define|true|[Yy])
+               # The fixed socket.h header file is wrong for gcc-4.x
+               # on PA-RISC2.0W, so Sock_type_t is size_t which is
+               # unsigned long which is 64bit which is too long
+               case "$gccversion" in
+                   4*) case "$archname" in
+                           PA-RISC*) socksizetype=int ;;
+                           esac
+                       ;;
+                   esac
+
                # For the moment, don't care that it ain't supported (yet)
                # by gcc (up to and including 2.95.3), cause it'll crash
                # anyway. Expect auto-detection of 64-bit enabled gcc on

diff --git a/pod/perldelta.pod b/pod/perldelta.pod
index 07f5dd3..0f83b03 100644 (file)
--- a/pod/perldelta.pod
+++ b/pod/perldelta.pod
@@ -339,6 +339,15 @@ XXX Any changes to platform support should be listed in the sections below.
 [ Within the sections, list each platform as a =item entry with specific
 changes as paragraphs below it. ]
 
+=over 4
+
+=item HP-UX PA-RISC/64 now supports gcc-4.x
+
+A fix to correct the socketsize now makes the test suite pass on HP-UX
+PA-RISC for 64bitall builds.
+
+=back
+
 =head2 New Platforms
 
 XXX List any platforms that this version of perl compiles on, that previous