patch gives no end of trouble when LFN=Y (I have numeric tails turned
OFF): witness:

C:\TEMP>set lfn=n

C:\TEMP>patch < a.pat
Hmm...  Looks like a new-style context diff to me...
The text leading up to this was:
--------------------------
|*** a.c        Wed Oct 23 07:29:30 1996
|--- b.c        Wed Oct 23 07:30:48 1996
--------------------------
Patching file a.c using Plan A...
Hunk #1 succeeded at 1.
done

C:\TEMP>set lfn=y

C:\TEMP>copy a.0 a.c
Overwrite a.c (Yes/No/All)?y
        1 file(s) copied

C:\TEMP>patch < a.pat
Hmm...  Looks like a new-style context diff to me...
The text leading up to this was:
--------------------------
|*** a.c        Wed Oct 23 07:29:30 1996
|--- b.c        Wed Oct 23 07:30:48 1996
--------------------------
Patching file a.c using Plan A...
Hunk #1 succeeded at 1.
Can't backup a.c, output is in c:/a/djgpp201/tmp/pofaaaaa: No such
file or directory (ENOENT)
done

C:\TEMP>patch -v
Patch version 2.1

C:\TEMP>dir c:\djgpp201in\patch.exe

 Volume in drive C is C @ WORK
 Volume Serial Number is 16D3-1278
 Directory of C:\A\djgpp201in

PATCH    EXE       117,248  10-19-96 11:28a patch.exe
         1 file(s)        117,248 bytes
         0 dir(s)      51,290,112 bytes free

C:\TEMP>ver

Windows 95. [Version 4.00.950]

I think this is a user/Win95 bug.  The user says he disabled numeric
tail.  Patch backs up `foo.bar' in a file called `~foo.bar'.  I
suspect that when numeric tail is disabled, Windows 95 cannot grok
such names.

FYI, Microsoft's official line is that when you disable numeric tails,
you are on your own, and they don't consider any strange effects of
this as their bugs.

Recompile with the latest release of libc.a in djdev201 (31st Oct 96).
patch.exe was compiled before libc's _rename.c file was corrected
for misbehaviour under Win95 with LFN=y.