Sender: rich AT phekda DOT freeserve DOT co DOT uk
Message-ID: <3E2BFD47.F2FCA284@phekda.freeserve.co.uk>
Date: Mon, 20 Jan 2003 13:44:39 +0000
From: Richard Dawe <rich AT phekda DOT freeserve DOT co DOT uk>
X-Mailer: Mozilla 4.77 [en] (X11; U; Linux 2.2.23 i586)
X-Accept-Language: de,fr
MIME-Version: 1.0
To: djgpp-workers AT delorie DOT com
Subject: Re: fstat, fd_props and inventing inodes [PATCH]
References: <E18XkXR-0001ZC-00 AT phekda DOT freeserve DOT co DOT uk> <2593-Wed15Jan2003095337+0200-eliz AT is DOT elta DOT co DOT il>
Content-Type: text/plain; charset=us-ascii
Content-Transfer-Encoding: 7bit
Reply-To: djgpp-workers AT delorie DOT com

Hello.

Eli Zaretskii wrote:
> 
> > Date: Sun, 12 Jan 2003 16:01:53 +0000
> > From: "Richard Dawe" <rich AT phekda DOT freeserve DOT co DOT uk>
[snip]
> > Basically: The patch uses fd_props to obtain the file name used
> > to open a file. It fixes up the file name and then uses
> > that as the input to _invent_inode.
> 
> Did you try to see what does this do with (a) handles open on devices,
> and (b) redirected handles?

Results for (a):

bash-2.04$ echo $CMD
./fstat.exe 0 /dev/null /dev/null con /dev/con c:/dev/con c:/dev/tty
bash-2.04$ $CMD
handle-0: -1 268435457 20644 1 42 0 1043068808 Mon Jan 20 13:20:08 2003
Everything checks out OK
handle-1: -1 268435457 20644 1 42 0 1043068808 Mon Jan 20 13:20:08 2003
Everything checks out OK
handle-2: -1 268435457 20644 1 42 0 1043068808 Mon Jan 20 13:20:08 2003
Everything checks out OK
handle-3: -1 268435458 20644 1 42 0 1043068808 Mon Jan 20 13:20:08 2003
Everything checks out OK
handle-4: -1 268435459 20644 1 42 0 1043068808 Mon Jan 20 13:20:08 2003
Everything checks out OK
/dev/null (7): -1 268435460 20644 1 42 0 1043068808 Mon Jan 20 13:20:08 2003
                        Times: 1043068808 312768000
                        Block size: 4096
Everything checks out OK
/dev/null (8): -1 268435460 20644 1 42 0 1043068808 Mon Jan 20 13:20:08 2003
                        Times: 1043068808 312768000
                        Block size: 4096
Everything checks out OK
con (9): -1 268435457 20644 1 42 0 1043068808 Mon Jan 20 13:20:08 2003
                        Times: 1043068808 312768000
                        Block size: 4096
Everything checks out OK
/dev/con (10): -1 268435457 20644 1 42 0 1043068808 Mon Jan 20 13:20:08 2003
                        Times: 1043068808 312768000
                        Block size: 4096
Everything checks out OK
c:/dev/con (11): -1 268435457 20644 1 42 0 1043068808 Mon Jan 20 13:20:08 2003
                        Times: 1043068808 312768000
                        Block size: 4096
Everything checks out OK
c:/dev/tty (12): -1 268435457 20644 1 42 0 1043068808 Mon Jan 20 13:20:08 2003
                        Times: 1043068808 312768000
                        Block size: 4096
Everything checks out OK

Test program for (b):

---Start fstat-redir.c---
#include <stdio.h>
#include <stdlib.h>
#include <time.h>
#include <sys/stat.h>

void
fstat_it (const char *argv0, const int fd)
{
  struct stat stat_buf;

  if (fstat(fd, &stat_buf) < 0)
    {
      perror(argv0);
      exit(EXIT_FAILURE);
    }

  /* Nicked from src/libc/posix/sys/stat/fstat.c */
  fprintf(stderr, "%s (%d): %d %6u %o %d %d %ld %lu %s", argv0, fd,
	  stat_buf.st_dev,
	  (unsigned)stat_buf.st_ino,
	  stat_buf.st_mode,
	  stat_buf.st_nlink,
	  stat_buf.st_uid,
	  (long)stat_buf.st_size,
	  (unsigned long)stat_buf.st_mtime,
	  ctime(&stat_buf.st_mtime));
  fprintf (stderr, "\t\t\tTimes: %lu %lu\n",
	   (unsigned long)stat_buf.st_atime,
	   (unsigned long)stat_buf.st_ctime);
  fprintf(stderr, "\t\t\tBlock size: %d\n",
	  stat_buf.st_blksize);
  _djstat_describe_lossage(stderr);
}

int
main (int argc, char *argv[])
{
  fstat_it(argv[0], fileno(stdin));
  fstat_it(argv[0], fileno(stdout));

  puts("wibble");

  return(EXIT_SUCCESS);
}
---End fstat-redir.c---

Results for (b):

bash-2.04$ rm -f foo foo2
bash-2.04$ echo blahblahblah > foo
bash-2.04$ ./fstat-redir <foo >foo2
c:/djgpp.204/rdawe-test/fstat/fstat-redir (0): 2 268435457 644 1 42 14
104307015
8 Mon Jan 20 13:42:38 2003
                        Times: 1043070158 1043070156
                        Block size: 16384
Failed to get starting cluster number; inode defaults to hashing
(if no other messages were printed, then this is either an empty
file on a local disk drive, or a file on a networked drive, or
you run under some kind of DOS clone)
SFT entry found, but is inconsistent with file size and time stamp
c:/djgpp.204/rdawe-test/fstat/fstat-redir (1): 2 268435458 644 1 42 0
1043070160
 Mon Jan 20 13:42:40 2003
                        Times: 1043070160 1043070158
                        Block size: 16384
Failed to get starting cluster number; inode defaults to hashing
(if no other messages were printed, then this is either an empty
file on a local disk drive, or a file on a networked drive, or
you run under some kind of DOS clone)
SFT entry found, but is inconsistent with file size and time stamp
bash-2.04$ ./fstat-redir
c:/djgpp.204/rdawe-test/fstat/fstat-redir (0): -1 268435457 20644 1 42 0
1043070
165 Mon Jan 20 13:42:45 2003
                        Times: 1043070165 312768000
                        Block size: 16384
Everything checks out OK
c:/djgpp.204/rdawe-test/fstat/fstat-redir (1): -1 268435457 20644 1 42 0
1043070
165 Mon Jan 20 13:42:45 2003
                        Times: 1043070165 312768000
                        Block size: 16384
Everything checks out OK
wibble

> Also, if you run _fixpath on a file name in a place other than where
> the file is actually opened, the program might be in a different
> directory, so _fixpath will produce incorrect results.  Therefore, I
> believe that if we want to use fd_props for this, we should run
> _fixpath on the file name when its info is recorded in fd_props.
> Then fstat should simply reuse the absolute name.

Well, it turns out that __set_fd_properties() runs the file name through
_truename(), before storing it. I think _truename() is sufficient - I don't
think we need to call _fixpath in fstat(). I think we can just use whatever
__get_fd_name() returns.

Bye, Rich =]

-- 
Richard Dawe [ http://www.phekda.freeserve.co.uk/richdawe/ ]