RE: [Gcl-devel] ANSI test Windows OPEN.* failures

[Mike Thomas]

Hi Camm.

| I see exactly the same thing you report below in Linux.
|
| cell-error-name might still just be a placeholder symbol.  See
| unixport/ansi_cl.lisp.
|
| error system still needs lots of work, but this should not be your
| problem at present, unless I misunderstand again!
|
| Please keep us posted.

These were indeed fruitful comments.

I decided last night to assume that the GCL error handling bugs are the same
on Linux as on Windows and that the real problem to solve in building Axiom
was not the error handling but the root cause of the error.

It turned out that the Spad compiler was unable to create a new file if the
file's proposed directory did not exist.  I solved this with
"ENSURE-DIRECTORIES-EXIST" in the function "get-io-index-stream" in
"src/interpreter/nlib.lisp.pamphlet" and when I left for work this morning
the spad compiler was happily doing algebra in the LAYER0COPY target in the
Makefile.

Why the Linux version of:

  (open index-file :direction :io :if-exists :overwrite
                       :if-does-not-exist :create)

apparently builds the directory automatically I must yet discover, but now
that I know where the error is, that should not be a problem.

I was using CVS GCL to build Axiom.

Cheers

Mike Thomas.