Re: NSK(OSS) compilation problem

[mwoehlke]

Paul Eggert wrote:
> mwoehlke <address@hidden> writes:
> > Ok... Attaching an initial attempt at a patch for stdint_.h. The
> > configure script (or more appropriately, I suppose, configure.in)
> > needs to be patched additionally, but alas, I do not speak autotools
> > :-)
>
> I'm afraid you'll have to get up to speed with them in order to do
> reasonable development in this area.  You don't have to port the
> autotools to NSK(OSS); just run them on a GNU/Linux host, do a "make
> dist" and then unpackage the tar image on NSK(OSS).

I was hoping someone else could slip in the 'configure needs to write a 
replacement for HAVE_UNSIGNED_LONG_LONG_INT'. :-) I would have thought 
this is a one-line, trivial change?

> > Anyway, this got me as far as sha512.c where the compiler croaked with:
>
> Fundamentally the problem is that sha512 assumes a 64-bit unsigned
> integer type, and you don't have one.  I guess the best way to fix
> this is to change coreutils to not build sha512 on hosts that don't
> have uint64_t.

Well, the question was "am I ok telling coreutils to not build 
sha512.c?". I'll take that as a "yes", thanks!

> > (So, yes... "yike"... Since this thread is already getting old (and
> > likely to get a lot older before I'm done), should I continue this
> > thread even if/when it starts falling off NTP, or start new ones?)
>
> I don't know what you mean by "NTP", but we're always interested in
> ports to real hosts, and Tandem/NSK OSS qualifies as a real host in my
> book.  The thread is still live on
> <http://lists.gnu.org/archive/html/bug-coreutils/2006-10/msg00062.html>,
> if that's what you mean.

Oops, I meant "NNTP"; I'm using gmane.org. What I meant was that I 
expect the roots of the post to start being culled, but I guess I'll 
just ignore that/live with it.

> > On an unrelated note, when running configure, I get these two header
> > compilation failures:
> >
> > ...
> > configure: WARNING: sys/statvfs.h: present but cannot be compiled
> > ...
> > configure: WARNING: sys/ioctl.h: present but cannot be compiled
> > ...
>
> This indicates that our tests for sys/ioctl.h and sys/statvfs.h are
> neglecting to include some prerequisite include files.  I cannot
> reproduce the problem with CVS coreutils on Solaris 10 (using Sun C
> 5.8), so I guess the problem is fixed now.  If not, you'll have to
> send us more info.

No, in *this* case (and it is a strange one), the problem is that 
_TANDEM_SOURCE is not defined; not by missing headers. I can verify this 
with:

$ CFLAGS=-D_TANDEM_SOURCE ./configure
...
checking sys/statvfs.h presence... yes
checking for sys/statvfs.h... yes
...
checking sys/ioctl.h usability... yes
checking sys/ioctl.h presence... yes
...

...but I don't feel that this is an ideal solution. I would prefer that 
autoconf know that if it is on NSK/OSS ('#ifdef __TANDEM'), that 
_TANDEM_SOURCE needs to be defined. I.e. I would add an early check that 
would toss '-D_TANDEM_SOURCE' into CFLAGS, and of course use that value 
as it goes on to try its test programs.

Meanwhile, however, I am going to keep going without to see what 
individual sources trip over this, unless I hear agreement that autoconf 
should be fixed. (I would guess this impacts quite a few packages, and 
not just coreutils.)

> > My guess, however, is that this is a case of
> > needing _TANDEM_SOURCE defined ala
> > http://savannah.gnu.org/bugs/?17172, but since this would need to be
> > changed in autoconf I'm not sure what, if anything, should or could be
> > done.
>
> Can you give more information about this?  What is _TANDEM_SOURCE
> used for?  You can check the files under /usr/include to find out.

If you see the aforementioned bug report, it must be defined for a 
number of headers to compile correctly... including sys/types.h (see 
next comment), and also sys/statvfs.h and sys/ioctl.h (see above).

> > For statvfs.h:
> > "/usr/include/sys/statvfs.h", line 50: error(114): identifier "u_long"
> > is undefined
>
> What include file is supposed to define u_long?  Use "grep -r u_long
> /usr/include" to find out.  Similary for ioctl.h.

sys/types.h defines u_*, but *only if _TANDEM_SOURCE is defined*. No 
header in /usr/include defines this, but *all* of these test for it:
/usr/include/ctype.h
/usr/include/dirent.h
/usr/include/fcntl.h
/usr/include/fnmatch.h
/usr/include/limits.h
/usr/include/locale.h
/usr/include/netdb.h
/usr/include/netinet/in.h
/usr/include/nl_types.h
/usr/include/pwd.h
/usr/include/signal.h
/usr/include/stdio.h
/usr/include/stdlib.h
/usr/include/string.h
/usr/include/stropts.h
/usr/include/sys/ipc.h
/usr/include/sys/socket.h
/usr/include/sys/types.h
/usr/include/sys/uio.h
/usr/include/sys/un.h
/usr/include/sys/wait.h
/usr/include/termios.h
/usr/include/unistd.h
/usr/include/wchar.h

I'm attaching sys/types.h so you can take a look, but just in case the 
horse isn't dead enough yet, /please/ just trust me that this *needs* to 
be defined :-).

> The patch you sent was incomplete.  I'm afraid I didn't really look at
> it much; instead, I constructed another patch from first principles,
> which I'll send in two later emails.

Thanks, I'll pull my changes to the affected files and try out your 
patch(es). If I am seeing correctly, however, aren't you missing patches 
to Makefile.in and configure (or rather sources there-to) so that 
@HAVE_UNSIGNED_LONG_LONG_INT@ is substituted correctly? Unless I am 
mistaken, without these the generated source file will wind up with 
literal '@HAVE_UNSIGNED_LONG_LONG_INT@' in it.

And thanks again for all the help; I really appreciate that you're 
willing to help support an exotic platform like this. :-)

--
Matthew
Will your shell have salvation? Only if it's Bourne Again.

#ifdef __TANDEM
#pragma columns 80
#if _TANDEM_ARCH_ != 0
#pragma ONCE
#endif
#endif
#ifndef _SYS_TYPES
#define _SYS_TYPES

/* T8645G08 - (16MAY2003) - sys/types.h   basic system types */

/*
 *  Copyright 2003 Compaq Computer Corporation
 *
 *     ALL RIGHTS RESERVED
 */

#ifdef __cplusplus
   extern "C" {
#endif


/*
 * In the OSS File System some of the definitions are 64-bit wide.
 * In Guardian C there is support of 64-bit wide variables declared
 * using long long.  To allow the same width on the workstation we will
 * use double.  The definition can be sourced in using the ifdef switch.
 */

#ifdef __TANDEM
#define int64_t long long
#else
typedef double  int64_t;
#endif

#ifndef __clock_t_DEFINED
#define __clock_t_DEFINED
typedef int64_t         clock_t;        /* Used for sys time in clock ticks */
#endif

typedef int64_t         dev_t;          /* <old device number> type */
typedef long            gid_t;          /* group id     */

#ifdef _XOPEN_SOURCE_EXTENDED
#ifndef __id_t_DEFINED
#define __id_t_DEFINED
typedef int             id_t;           /* Can hold gid_t, pid_t, or uid_t */
#endif
#endif /* _XOPEN_SOURCE_EXTENDED */

typedef long            ino_t;          /* <inode> type */
typedef int             key_t;          /* IPC Key */
typedef unsigned long   mode_t;
typedef short           nlink_t;

#ifndef __off_t_DEFINED
#define __off_t_DEFINED
typedef long            off_t;          /* file offset (signed number) */
#endif

typedef int             pid_t;          /* process id (signed number) */

#ifndef __size_t_DEFINED
#define __size_t_DEFINED
#if _TANDEM_ARCH_ != 0 || !defined(__TANDEM)
typedef unsigned int   size_t;
#else /* _TANDEM_ARCH_ == 0 */
#if !defined(_GUARDIAN_TARGET) || defined(__XMEM)
typedef unsigned long  size_t;
#else
typedef unsigned short size_t;
#endif
#endif /* _TANDEM_ARCH_ == 0 */
#endif /* __size_t_DEFINED */

#ifndef __ssize_t_DEFINED
#define __ssize_t_DEFINED
#if _TANDEM_ARCH_ != 0 || !defined(__TANDEM)
typedef int   ssize_t;
#else /* _TANDEM_ARCH_ == 0 */
#if !defined(_GUARDIAN_TARGET) || defined(__XMEM)
typedef long  ssize_t;
#else
typedef short ssize_t;
#endif
#endif /* _TANDEM_ARCH_ == 0 */
#endif /* __ssize_t_DEFINED */

#ifndef __time_t_DEFINED
#define __time_t_DEFINED
typedef long            time_t;         /* time in seconds */
#endif

typedef long            uid_t;          /* user id      */

#ifdef _TANDEM_SOURCE
#ifndef __caddr_t_DEFINED
#define __caddr_t_DEFINED
typedef char   *caddr_t;
#endif
typedef unsigned char   uchar_t;
typedef unsigned int    uint_t;

#ifndef _tcp_TYPES
#define _tcp_TYPES
typedef unsigned char  u_char;
typedef unsigned short u_short;
typedef unsigned int   u_int;
typedef unsigned long  u_long;
typedef unsigned short ushort;
#endif /* _tcp_TYPES */

#ifndef __daddr_t_DEFINED
#define __daddr_t_DEFINED
typedef long    daddr_t;
#endif

#ifndef __NBBY_DEFINED
#define __NBBY_DEFINED
#define NBBY    8               /* number of bits in a byte */
#endif

#ifndef howmany
#define howmany(x, y)   (((x)+((y)-1))/(y))
#endif
#endif /* _TANDEM_SOURCE */


#ifdef __cplusplus
   }
#endif

#endif  /* _SYS_TYPES defined */