[M3commit] CVS Update: cm3

[Tony Hosking]

Darwin no longer uses wrapper functions for pread etc.

I figured a static switch at build time.

On 4 Jan 2009, at 19:35, Jay wrote:

> Exactly how?
> Like off_t == word?
> That isn't right.
>
> Switch which files?
> I want to reduce switching more than what is needed.
> Can the string in an extern be "forwarded"?
> That would be the ideal granularity here.
> Darwin does stuff very similar or maybe the same as I am doing --  
> wrapper functions in C for reasons related to off_t, though the  
> wrappers are a bit dubious, e.g.:
>
> int m3_pread(int d, void *buf, size_t nbytes, int offset)
> {
>   off_t off = (off_t)offset;
>   return pread(d, buf, nbytes, offset);
> }
>
> int m3_profil(char *samples, size_t size, u_long offset, u_int scale)
> {
>   off_t off = (off_t)offset;
>   return profil(samples, size, offset, scale);
> }
>
> int m3_pwrite(int d, const void *buf, size_t nbytes, int offset)
> {
>   off_t off = (off_t)offset;
>   return pwrite(d, buf, nbytes, offset);
> }
>
> int m3_ftruncate(int fd, int length)
> {
>   off_t len = (off_t) length;
>   return ftruncate(fd, length);
> }
>
> What does the first line of each of those accomplish?
>
> void
> m3_assignOffT (off_t* dest, long src)
> {
>   *dest = (off_t)src;
> }
>
> Is that widening or narrowing, and if it is narrowing, silent  
> truncation?
> Granted, Modula-3 doesn't check for integer overflow anyway, so  
> silent truncation I guess is "ok".
>
> Ideally, every system would have a 64 bit off_t, if not the default,  
> at least via #defines, or using function names with "64" appended  
> (or both). The case for 32bit off_t would be interop with existing  
> code. Otherwise, it seems clear that it should always, if possible,  
> be 64 bits. time_t should be 64 bits too, but that's another matter.
>
> I did just check and DJGPP seems limited to 32bit off_t.
> So eventually it might have to go back to Usysdep.i3, but ok.
>
>  - Jay
>
>
>
> From: hosking at cs.purdue.edu
> To: jay.krell at cornell.edu
> Date: Sun, 4 Jan 2009 15:54:48 +1100
> CC: m3commit at elegosoft.com
> Subject: Re: [M3commit] CVS Update: cm3
>
> Why not just switch between 32-bit and 64-bit versions.
>
> Antony Hosking | Associate Professor | Computer Science | Purdue  
> University
> 305 N. University Street | West Lafayette | IN 47907 | USA
> Office +1 765 494 6001 | Mobile +1 765 427 5484
>
>
>
> On 4 Jan 2009, at 13:22, Jay wrote:
>
> But how about if we always #define _FILE_OFFSET_BITS 64 and use C  
> wrappers?
>   As I have done.
> Or externs with "64" appended, but that prevents me from using one  
> set of .i3 files.
>
> I mean, I know there are "modes" with 32 bit off_t, but we should  
> really strive not to use them, if given a choice, eh?
>
> You think there are any systems we will ever care about that don't  
> offer a 64 bit off_t (or off64_t, or whatever).
>
>  - Jay
>
>
> From: hosking@
> To: jkrell@
> Date: Sun, 4 Jan 2009 12:57:28 +1100
> CC: m3commit at elegosoft.com
> Subject: Re: [M3commit] CVS Update: cm3
>
> On some platforms off_t is not 64 bits.  This from Solaris (see  
> off_t==long):
>
> #if defined(_LARGEFILE_SOURCE) || defined(_XPG5)
> #ifndef _OFF_T
> #define _OFF_T
> #if defined(_LP64) || _FILE_OFFSET_BITS == 32
> typedef long            off_t;
> #else
> typedef __longlong_t    off_t;
> #endif
> #ifdef  _LARGEFILE64_SOURCE
> #ifdef _LP64
> typedef off_t           off64_t;
> #else
> typedef __longlong_t    off64_t;
> #endif
> #endif /* _LARGEFILE64_SOURCE */
> #endif /* _OFF_T */
> #endif /* _LARGEFILE_SOURCE *
>
>
> Antony Hosking | Associate Professor | Computer Science | Purdue  
> University
> 305 N. University Street | West Lafayette | IN 47907 | USA
> Office +1 765 494 6001 | Mobile +1 765 427 5484
>
>
>
> On 4 Jan 2009, at 01:53, Jay Krell wrote:
>
> CVSROOT: /usr/cvs
> Changes by: jkrell at birch. 09/01/04 01:53:03
>
> Modified files:
> cm3/m3-libs/m3core/src/unix/Common/: Unix.i3 Utypes.i3
> cm3/m3-libs/m3core/src/unix/cygwin/: Usysdep.i3
> cm3/m3-libs/m3core/src/unix/linux-common/: Usysdep.i3
> cm3/m3-libs/m3core/src/unix/openbsd-common/: Usysdep.i3
>
> Log message:
> significant reduction in sysdep content
>
> off_t is always 64 bits (or at least let's hope so)
> the dir/stat flags are mostly all the same,
> except for a few that are zero on cygwin
>
> M[RWX][OWNER GROUP OTHER]
> [FXWR]_OK
> F_SETFD, F_GETFL, F_SETFL
> FIONREAD
> ITIMER_REAL
> ITIMER_VIRTUAL
>
> Some of this might be by chance and not by-standard,
> so under review for new platforms, and probably should be verified
> during native builds.
>
> note that OS/2 has a different S_IFMT but not quite relevant here
>
>
>
>

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://m3lists.elegosoft.com/pipermail/m3commit/attachments/20090104/110a58e4/attachment-0002.html>