irix6-libc-compat.c: New file.

* config/mips/irix6-libc-compat.c: New file.
	* config/mips/t-iris6 (LIB2FUNCS_STATIC_EXTRA): Use it.
	* doc/install.texi (Specific, mips*-sgi-irix6): Mention structure
	passing workaround.

From-SVN: r44294

gcc/ChangeLog

@@ -1,3 +1,10 @@
+2001-07-24  Rainer Orth  <ro@TechFak.Uni-Bielefeld.DE>
+
+	* config/mips/irix6-libc-compat.c: New file.
+	* config/mips/t-iris6 (LIB2FUNCS_STATIC_EXTRA): Use it.
+	* doc/install.texi (Specific, mips*-sgi-irix6): Mention structure
+	passing workaround.
+
 2001-07-24  lars brinkhoff  <lars@nocrew.org>
 
 	* rtl.texi (REG_POINTER): Document.

gcc/config/mips/irix6-libc-compat.c

@@ -0,0 +1,140 @@
+/* Compensate for inconsistent structure passing conventions on IRIX 6.  */
+/* Compile this one with gcc.  */
+/* Copyright (C) 2001  Free Software Foundation, Inc.
+
+This file is part of GNU CC.
+
+GNU CC is free software; you can redistribute it and/or modify
+it under the terms of the GNU General Public License as published by
+the Free Software Foundation; either version 2, or (at your option)
+any later version.
+
+In addition to the permissions in the GNU General Public License, the
+Free Software Foundation gives you unlimited permission to link the
+compiled version of this file into combinations with other programs,
+and to distribute those combinations without any restriction coming
+from the use of this file.  (The General Public License restrictions
+do apply in other respects; for example, they cover modification of
+the file, and distribution when not linked into a combine
+executable.)
+
+GNU CC is distributed in the hope that it will be useful,
+but WITHOUT ANY WARRANTY; without even the implied warranty of
+MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+GNU General Public License for more details.
+
+You should have received a copy of the GNU General Public License
+along with GNU CC; see the file COPYING.  If not, write to
+the Free Software Foundation, 59 Temple Place - Suite 330,
+Boston, MA 02111-1307, USA.  */
+
+/* GCC doesn't correctly implement the structure and union passing and return
+   conventions of the N32 and N64 ABIs on IRIX 6, as described in the
+   MIPSpro N32 ABI Handbook, ch. 2, Calling Convention Implementations, p.7.
+   The ABI requires that structures (or trailing parts of structures) smaller
+   than 8 bytes (a 64-bit register) are left-justified, whereas GCC
+   right-justifies them.
+
+   While GCC is internally consistent, calling routines compiled with a
+   compiler that does implement the documented ABI (like SGIs MIPSpro C
+   compiler) doesn't work.  This is primarily an issue for system libraries
+   like libc.  Fortunately, there exist only very few routines that take
+   structure value arguments or return structures by value, so until the
+   underlying bug is fixed, it is possible to work around it by providing
+   wrapper functions for the few affected routines that compensate for the
+   inconsistent alignment.
+
+   These wrappers rely on the fact that e.g. libc contains weak versions of
+   those routines, and the real implementation is provided by _-prefixed
+   variants.  So we can provide our own versions, which will only be linked
+   if the application uses any of the affected functions, calling the private
+   variants after shifting the arguments or results as required.
+
+   This is a rewrite of code created by Andy Polyakov.  */
+
+#include "config.h"
+#include "system.h"
+
+/* This must only be used for the N32 and N64 ABIs.  O32 is correct.  */
+
+#if _MIPS_SIM == _ABIN32 || _MIPS_SIM == _ABI64
+
+/* The affected arguments need to be shifted by
+
+	BITS_PER_WORD - (sizeof (arg) * BITS_PER_UNIT).
+
+   Since only 32-bit args and results are involved, the shift count is
+   always 32.  */
+#define SHIFT_BITS	32
+
+extern machreg_t _inet_ntoa	PARAMS ((machreg_t));
+extern machreg_t _inet_lnaof	PARAMS ((machreg_t));
+extern machreg_t _inet_netof	PARAMS ((machreg_t));
+extern machreg_t _inet_makeaddr	PARAMS ((machreg_t, machreg_t));
+
+/* <arpa/inet.h> has
+
+	char *inet_ntoa (struct in_addr);
+
+   on both IRIX 6.2 and 6.5, with struct in_addr containing a 32-bit int.  */
+
+machreg_t
+inet_ntoa (machreg_t in)
+{
+  return _inet_ntoa (in << SHIFT_BITS);
+}
+
+/* <arpa/inet.h> has
+
+	unsigned long inet_lnaof (struct in_addr);		(IRIX 6.2)
+	in_addr_t inet_lnaof (struct in_addr);			(IRIX 6.5)
+
+   in_addr_t is a 32-bit int.  */
+
+machreg_t
+inet_lnaof (machreg_t in)
+{
+  return _inet_lnaof (in << SHIFT_BITS);
+}
+
+/* <arpa/inet.h> has
+
+	unsigned long inet_netof (struct in_addr);		(IRIX 6.2)
+	in_addr_t inet_netof (struct in_addr);			(IRIX 6.5)  */
+
+machreg_t
+inet_netof (machreg_t in)
+{
+  return _inet_netof (in << SHIFT_BITS);
+}
+
+/* <arpa/inet.h> has
+
+	struct in_addr inet_makeaddr (int, int);		(IRIX 6.2)
+	struct in_addr inet_makeaddr (in_addr_t, in_addr_t);	(IRIX 6.5)  */
+
+machreg_t
+inet_makeaddr (machreg_t net, machreg_t lna)
+{
+  return _inet_makeaddr (net, lna) >> SHIFT_BITS;
+}
+
+#if _MIPS_SIM == _ABIN32
+extern machreg_t _semctl	PARAMS ((machreg_t, machreg_t, machreg_t, machreg_t));
+
+/* <sys/sem.h> has
+
+	int semctl (int, int, int, ...);
+
+   where the variadic argument is union semun if used.  union semun contains
+   an int and two pointers, so the union is already 64 bits wide under the
+   N64 ABI and alignment is not an issue.  */
+
+machreg_t
+semctl (machreg_t semid, machreg_t semnum, machreg_t cmd, machreg_t arg)
+{
+  return _semctl(semid, semnum, cmd, arg << SHIFT_BITS);
+}
+#endif /* _ABIN32 */
+
+#endif /* _ABIN32 || _ABI64 */

gcc/config/mips/t-iris6

@@ -53,3 +53,7 @@ fp-bit.c: $(srcdir)/config/fp-bit.c
 	echo '#define FLOAT' > fp-bit.c
 	echo '#undef US_SOFTWARE_GOFAST' >> fp-bit.c
 	cat $(srcdir)/config/fp-bit.c >> fp-bit.c
+
+# This is only needed in the static libgcc as a band-aid until gcc correctly
+# implements the N32/N64 ABI structure passing conventions
+LIB2FUNCS_STATIC_EXTRA = $(srcdir)/config/mips/irix6-libc-compat.c

gcc/doc/install.texi

@@ -2779,7 +2779,8 @@ GCC is consistent with itself, but not consistent with the SGI C compiler
 happen are when there are library functions that take/return such
 structures. There are very few such library functions.  Currently this
 is known to affect @code{inet_ntoa}, @code{inet_lnaof},
-@code{inet_netof}, @code{inet_makeaddr}, and @code{semctl}.
+@code{inet_netof}, @code{inet_makeaddr}, and @code{semctl}.  Until the
+bug is fixed, GCC contains workarounds for the known affected functions.
 
 See @uref{http://freeware.sgi.com/,,http://freeware.sgi.com/} for more
 information about using GCC on IRIX platforms.