Re: bootstrap/3163

Re: bootstrap/3163

[Robert Lipe]

The following reply was made to PR bootstrap/3163; it has been noted by GNATS.

From: Robert Lipe <robertlipe@usa.net>
To: gcc-gnats@gcc.gnu.org, gcc-prs@gcc.gnu.org, john@Calva.COM,
        gcc-bugs@gcc.gnu.org, nobody@gcc.gnu.org
Cc:  
Subject: Re: bootstrap/3163
Date: Wed, 13 Jun 2001 11:45:25 -0500

 http://gcc.gnu.org/cgi-bin/gnatsweb.pl?cmd=view&pr=3163&database=gcc
 
                HP/UX allocates pages backward in memory. No OS has yet
 been
                observed
                to be so perverse as to leave unmapped space between
                consecutive calls
                to mmap. */
 
                Well, UnixWare 7.1.1 is that perverse system. It seems to
                leave
                one page between each mapping.
 
 This is actually documented behaviour.   From the mmap man page:
 
    References to pages in a mapped region which are beyond the
    page containing the end of a file will result in the delivery of a
    SIGBUS signal.
 
 It sounds like this test is looking for several undocumented 
 characteristics of mmap that just happen to be true on some systems
 and not a general case mmap.  Since the code provides a reasonable
 fallback, does this cause an actual problem or is it an observation
 that this test is failing becuase the OS doesn't satisfy the 
 characteristics wanted by this test?

RE: bootstrap/3163

[John Hughes]

The following reply was made to PR bootstrap/3163; it has been noted by GNATS.

From: "John Hughes" <john@Calva.COM>
To: "Robert Lipe" <robertlipe@usa.net>, <gcc-gnats@gcc.gnu.org>,
   <gcc-prs@gcc.gnu.org>, <gcc-bugs@gcc.gnu.org>, <nobody@gcc.gnu.org>
Cc:  
Subject: RE: bootstrap/3163
Date: Thu, 14 Jun 2001 09:28:32 +0200

 >> http://gcc.gnu.org/cgi-bin/gnatsweb.pl?cmd=view&pr=3163&database=gcc
 >> No OS has yet been observed to be so perverse as to leave unmapped
 >> space between consecutive calls to mmap. */
 >>
 >> Well, UnixWare 7.1.1 is that perverse system. It seems to leave
 >> one page between each mapping.
 
 > This is actually documented behaviour.   From the mmap man page:
 >
 >   References to pages in a mapped region which are beyond the
 >   page containing the end of a file will result in the delivery of a
 >   SIGBUS signal.
 
 No, that paragraph says that if I mmap 10 pages from a 5 page file
 access to the 5 pages beyond eof get a SIGBUS (note: not SIGSEGV).
 
 It says nothing about memory OUTSIDE the mmaped region.
 

Re: bootstrap/3163

[Zack Weinberg]

The following reply was made to PR bootstrap/3163; it has been noted by GNATS.

From: "Zack Weinberg" <zackw@stanford.edu>
To: Robert Lipe <robertlipe@usa.net>
Cc: gcc-gnats@gcc.gnu.org, john@Calva.COM
Subject: Re: bootstrap/3163
Date: Wed, 13 Jun 2001 20:33:38 -0700

 On Wed, Jun 13, 2001 at 11:45:25AM -0500, Robert Lipe wrote:
 > http://gcc.gnu.org/cgi-bin/gnatsweb.pl?cmd=view&pr=3163&database=gcc
 > 
 >                HP/UX allocates pages backward in memory. No OS has yet
 > been
 >                observed
 >                to be so perverse as to leave unmapped space between
 >                consecutive calls
 >                to mmap. */
 > 
 >                Well, UnixWare 7.1.1 is that perverse system. It seems to
 >                leave
 >                one page between each mapping.
 > 
 > This is actually documented behaviour.   From the mmap man page:
 > 
 >    References to pages in a mapped region which are beyond the
 >    page containing the end of a file will result in the delivery of a
 >    SIGBUS signal.
 
 That sentence is in everybody's mmap(2), but not everybody's mmap
 leaves one page between each successive anonymous mapping.
 
 The test is, however, too picky.  I originally wrote it to exercise
 every last detail of a Cygwin runtime bug.  I've seen problem reports
 go by for it on Solaris 8 and the Hurd as well as UnixWare 7.1.1.
 Fixing it just hasn't been very high on my list.
 
 It would probably be better to check for the buggy versions of Cygwin
 instead.  I don't know how to do that reliably, or which version is
 the earliest in which the bug was fixed.
 
 See thread at http://gcc.gnu.org/ml/gcc-bugs/2000-11/msg00588.html .
 
 -- 
 zw            I now propose to do a little a priorizing of my own.
               	-- Jerry Fodor, _The Mind Doesn't Work That Way_