JDBP>> The example that I sent them was the behaviour of `FFIND \*S' as
 JDBP>> compared to `DIR \*S',...
 JDBP>> The latter group matches the file `\CONFIG.SYS' and doesn't match
 JDBP>> the file `\DOS.DAT'.  This is, as I'm sure everyone here reading
 JDBP>> this will agree, eminently correct.  The FFIND command, however,
 JDBP>> operates in the reverse manner.  `FFIND \*S' matches the file
 JDBP>> `\DOS.DAT' and *doesn't* match the file `\CONFIG.SYS'.

 KM> The matter under 4DOS here is different from what you obviously expect
 KM> and very complex:

I thought I made this very clear in my original post.  Obviously not.  I'll
repeat it:

JdeBP> But I hadn't been talking about DOS.  I was in fact referring to 
JdeBP> all of their non-DOS command interpreters (and I had tested most 
JdeBP> of them and found this bug in FFIND everywhere that I tested, and 
JdeBP> had explicitly listed, in my original bug report, which ones I 
JdeBP> had tested and on what platforms).  

So there's no point in your lecturing me on how wildcard matching in 4DOS
works.  I *know* how wildcard matching in 4DOS works.  The point is that by
design IN NO OTHER OPERATING SYSTEM DOES WILDCARD MATCHING WORK IN THE SAME
WAY AS IT DOES IN DOS.  Wildcard matching by the DOS COMMAND, and hence by
4DOS since it has to retain compatibility, is generally considered to be
broken, and most other operating systems use what are generally agreed to
be far saner semantics -- that "*S" matches any name whose *last*
character is 'S', for example.  This *doesn't* include the name
"DOS.DAT" and *does* include "CONFIG.SYS", yet in 4OS2,
Take Command for OS/2, 4NT, and Take Command for Win32 the command
"FFIND \*S" *will* match the former and *will not* match the
latter.  

Given, firstly, that DEL, MOVE, COPY, REN, and all of the other commands
that take wildcards in the very same command interpreters, all use the
correct semantics, and secondly, that it is blindingly obvious just from
looking at the wildcard "*S" that (on an operating system with
sane wildcard semantics such as OS/2 Warp or Windows NT)
"DOS.DAT" *shouldn't* match and "CONFIG.SYS" *should*,
it is clear that FFIND has a bug here.  Yet JP Software Technical Support
refuses to accept this.

 KM> So part of the problem might come from emulating CC's behaviour.

The behaviour of COMMAND is irrelevant, which should be clear from
re-reading the part of my original post that is quoted above.  

 ¯ JdeBP ®

--- FleetStreet 1.19 NR