Hi there,

I've been doing some work on the version of CSPLIT Jasen Betts posted
and its path save/re-creation, and as a result have some points I'd
like your thoughts on.

1) Saveing the path information isn't difficult, but as it is stored
   togther with the file name in a header line and there are no checks
   on header line lengths it can create header lines that are longer
   than the specified maximum line length.

   I suspect this is not acceptable, views please.

2) Saveing the path information as Jasen has means the files are not
   compatible with older versions of CSPLIT if the source files were not
   in the current directory.

   I think that this is not acceptable, again your comments.

3) Restoring the paths requires use of mkdir(), which is not ANSI
   although it is commonly available in PC compilers. I've structured
   the code to check __STDC__ and if __STDC__ is defined it will not
   store path information, will not support "-xd" on extraction, and
   will dump the path information if any is present.

   Should it always store path information (but see above notes)
   which is possible within ANSI, or is this OK?

4) Given the can of worms on path saveing/re-creation should we go back
   the older CSPLIT 2.2 approach, and never store path information?

I have looked at using the SNIPPETS mkdirs() function, but this is still
not a cleanly portable solution (under DOS it ends up using an assembler
module and DOS interrupts) so I stuck with mkdir().

Aside to Bob Stout: During this I've updated one of the SNIPPETS header
files to handle the Borland OS/2 compiler, when things have settled down
I'll pass it on.

George

 * SLMR 2.1a * KEYBOARD - Used to enter errors int a computer.

--- Maximus/2 3.01