Project 15: Non-numerical aids for numerical testing

Oxford 1975:
============
 
It was agreed that Dr Einarsson should collect further information 
on non-numerical aids for numerical testing.

Boulder 1981:
=============

Fosdick reviewed the progress of Toolpack and provided a list of
documents (IFIP/WG2.5 (Boulder-27) 827). Any of the documents listed
are available from Wayne Cowell. Also anyone wishing to become a
"friend of Toolpack" should tell him. Release -1 of Toolpack is an
internal release scheduled for September 30, 1981.  It consists of
only about 50 % of tools finally planned and they are not in their
final form. F&ease 0 is planned for May 1982 to special friends of
Toolpack.  Release 1 will be for productive use and will appear later.
Release -1 includes a lexer, parser, flow analyzer, formatter, editor,
call grapher, structurer, template (as in Ellpack), a recast of Unix
facilities, command language system, master file system, input/output
primitives to avoid Fortran i/o problems. Toolpack is written in any
language that can be pre-processed into PFORT. It will accept an
extension (as wanted 8 in various labs) of Fortran 77. It is not a
design aim to accept Fortran 66 and some incompatibilities may cause
problems. Brown thought that a good goal would be to accept both
Fortrans, but Reinsch supported the F-77 decision. In reply to
questions from Reinsch, Fosdick said that dynamic tracing will be
provided by insertions of source statements, that a macro processor,
more powerful than AUGMENT, will be built. There would be no
interactive syntax scanner or interactive interpretation, though some
facilities will be available from the Fortran intelligent editor.

Lawson commented on the relationship of Toolpack to its original aims.
It was found that different tools did not work together so a command
language was needed to pull them together. Now this aspect appears to
dominate and the "tool parts" concept has been de-emphasized.  Fosdick
felt that this may not matter.

Reid asked whether there would be a verifier for Fortran 77. Fosdick
said that this would be part of the flow analyzer. Brown commented on
the need for a "destructurer" to complement the structurer.

Madison 1982:
=============

Ford reported on the progress of the Toolpack project.  Release -1 was
sent only to the Toolpack group on November 15, 1981 and consisted of
a few representative tools and tool fragments. Release 0 is intended
to be sent in September 1982 to the toolpack group and 7 selected test
sites. It will consist of a command interpreter and a significant set
of tools (some in prototype form). Release 1 is aimed for availability
to an extended audience (perhaps including the public) in summer or
autumn 1983 and will contain a first release of all intended major
tools of the IST (integrated system of tools). The Toolpack advisory
board has been formed and had its first meeting in Boulder in February
1982. The "friends of Toolpack" now include 300 from Europe.

Lawson remarked that the project is more difficult than EISPACK or
LINPACK. The JPL people (Kroch, Schneider) are critical of the work.
They want low-level things to be built first and the rest built on top
and will probably separate from Toolpack. Einarsson recommended being
a friend of Toolpack; it meant the receipt, at least once a year, of a
few pages of information. Ford offered to make copies of the
documentation for those interested. A majority of the group requested
further information.

Argonne 1986:
=============

B. Ford presented an overview and current status of the Toolpack project. The first 
public release of Toolpack (version 1) was made in February 1985. The second release 
is due towards the end of 1986. The second release had benefited from continued 
development and the exposure of the first release to a wide audience. Some of the 
problems, such as slowness of the system, and installation difficulties in IBM 
environments were pointed out and corrective actions will be taken wherever
possible. Discussants noted that distribution of the system in object form (already 
tailored to a particular environment) may be desirable. Possibilities of toolpack-like 
systems for other languages, separate vs. integrated tools, and TIE interface were also 
discussed.