From: Martin Str|mberg <ams AT ludd DOT luth DOT se>
Message-Id: <199811021736.SAA17007@father.ludd.luth.se>
Subject: Re: df adventures
To: eliz AT is DOT elta DOT co DOT il (Eli Zaretskii)
Date: Mon, 2 Nov 1998 18:36:08 +0100 (MET)
Cc: djgpp-workers AT delorie DOT com (DJGPP-WORKERS)
In-Reply-To: <Pine.SUN.3.91.981102095037.11234A-100000@is> from Eli Zaretskii at "Nov 2, 98 09:51:11 am"
X-Mailer: ELM [version 2.4ME+ PL15 (25)]
MIME-Version: 1.0
Content-Type: text/plain; charset=US-ASCII
Content-Transfer-Encoding: 7bit
Reply-To: djgpp-workers AT delorie DOT com

According to Eli Zaretskii:
> I still would like to know what causes that "No such drive" error, and
> whether "df r:" crashes or not after this patch (if it does crash, it
> seems like a bug in `df', judging by the traceback).
> 
> *** src/libc/compat/mntent/mntent.c~5	Mon Oct 12 23:22:50 1998
> --- src/libc/compat/mntent/mntent.c	Mon Nov  2 09:27:10 1998
> *************** getmntent(FILE *filep)
> *** 787,796 ****
>             /* Floppies and other disks written by Windows 9X include
>                entries that have volume label bit set, but they are
>                actually parts of some LFN entry.  We only accept volume
> !              labels which have all their RHS bits reset; otherwise
> !              we pretend we never saw that label.  */
>             while (volume_found &&
> !               (mnt_ff.ff_attrib & (FA_RDONLY|FA_HIDDEN|FA_SYSTEM)) != 0)
>               {
>                 volume_found = 0;
>                 errno = ENMFILE;
> --- 787,796 ----
>             /* Floppies and other disks written by Windows 9X include
>                entries that have volume label bit set, but they are
>                actually parts of some LFN entry.  We only accept volume
> !              labels which have their HS bits reset; otherwise we
> !              pretend we never saw that label.  */
>             while (volume_found &&
> ! 		 (mnt_ff.ff_attrib & (FA_HIDDEN|FA_SYSTEM)) != 0)
>               {
>                 volume_found = 0;
>                 errno = ENMFILE;
> 

This patch makes "df" work, i. e. R: is seen under condition that
there is a CDROM with data in the drive. "df r:" works fine too in
this case. 

However running "df r:" with no disk in the CDROM drive results in a
crash: 

E:\>df3.exe r:
Filesystem         1024-blocks  Used Available Capacity Mounted on
Exiting due to signal SIGSEGV
Page fault at eip=00001ab2, error=0004
eax=00000000 ebx=0009a2f8 ecx=00000005 edx=00000001 esi=00000000 edi=000019d7
ebp=00099aa0 esp=00099a00 program=E:\DF3.EXE
cs: sel=00a7  base=10000000  limit=0009ffff
ds: sel=00af  base=10000000  limit=0009ffff
es: sel=00af  base=10000000  limit=0009ffff
fs: sel=008f  base=000064d0  limit=0000ffff
gs: sel=00bf  base=00000000  limit=0010ffff
ss: sel=00af  base=10000000  limit=0009ffff
App stack: [00099b40..00019b40]  Exceptn stack: [00019a10..00017ad0]

Call frame traceback EIPs:
  0x00001ab2   _show_dev+78, line 273 of df.c
  0x00001ff4   _show_point+168, line 512 of df.c
  0x00002057   _show_entry+47, line 529 of df.c
  0x00002cee   _main+1554, line 779 of df.c
  0x0000655e   ___crt1_startup+174

The problem is this line in df.c:

  /* We couldn't find the mount entry corresponding to POINT.  Go ahead and
     print as much info as we can; methods that require the device to be
     present will fail at a later point.  */
  {
    /* Find the actual mount point.  */
    char *mp = find_mount_point (point, statp);
    if (mp)
      {
        show_dev (0, mp, 0);
^^^^^^^^^^^^^^^^^^^^^^^^^^^^
        free (mp);
      }
    else
      error (0, errno, "%s", point);
  }

In show_dev there's a strcmp using the third parameter:

static void
show_dev (const char *disk, const char *mount_point, const char *fstype)
{
...
  /* JOINed disks seem like a directory on another disk.
     We want to see the info about the real drive.  */
  if (strcmp (fstype, "join") == 0)
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
    {
      struct stat real_drive_stats;

So one solution would be changing the first line to "show_dev (0, mp,
"");"  Another one would be changing the second one to "if (fstype &&
strcmp (fstype, "join") == 0)". There's perhaps more of them. But I
don't know what is the Right Way(tm) to do it. Perhaps Eli has the
right answer?

I'll ignore the "df: f:/: No such device (ENODEV)" trouble as this
started when I have two primary FAT16 partitions on one (and only)
hard drive. *DOZE was never intended to support this...


Right,

							MartinS