From mboxrd@z Thu Jan 1 00:00:00 1970 From: ro@gcc.gnu.org To: ake@purdue.edu, gcc-bugs@gcc.gnu.org, gcc-prs@gcc.gnu.org, java-prs@gcc.gnu.org, nobody@gcc.gnu.org, ro@gcc.gnu.org Subject: Re: libgcj/3694: natPlainDatagramSocketImpl.cc doesn't compile in ALPHA OSF1 Date: Thu, 26 Jul 2001 12:00:00 -0000 Message-id: <20010726190037.4996.qmail@sourceware.cygnus.com> X-SW-Source: 2001-07/msg00719.html List-Id: Synopsis: natPlainDatagramSocketImpl.cc doesn't compile in ALPHA OSF1 Responsible-Changed-From-To: unassigned->ro Responsible-Changed-By: ro Responsible-Changed-When: Thu Jul 26 12:00:36 2001 Responsible-Changed-Why: Mine. State-Changed-From-To: open->analyzed State-Changed-By: ro State-Changed-When: Thu Jul 26 12:00:36 2001 State-Changed-Why: The socklen_t problem also exists on Tru64 UNIX V4.0F with patch kit 2, but not with patch kit 4, and is also fixed on V5.1. I on V4.0F with patch kit 4, the definition of socklen_t is conditional on the definition of _POSIX_PII_SOCKETS, the same feature test macro used to select either the new (posix-style with socklen_t) or old (with size_t) declarations of various socket functions. If a similar patch exists for V5.0 and V5.0A, this could be used as a workaround. I tried to define _POSIX_PII_SOCKETS myself in java/net/natPlainDatagramSocketImpl.cc and java/net/natPlainSocketImpl.cc, but this triggers a different problem: with this macro defined, accept is changed into a three-argument macro, conflicting with the one-argument accept in java/net/PlainSocketImpl.h (from PlainSocketImpl.java). We cannot work around this by replacing accept by e.g. _accept, since this accept method is part of the public interface. Ideas? I don't know what to do about the mreq6 problem, since I don't have any V5.0/V5.0A machine. But I recall that while IPv6 was first introduced in V5.1, V5.0 had some of the types etc. mandated by the IPv6 extensions to the socket api. http://gcc.gnu.org/cgi-bin/gnatsweb.pl?cmd=view&pr=3694&database=gcc