The GCC Cobol project seems to be maturing into a beta version. Much
of the work since our last posting in November 2023 has been devoted to
problem reports from users. The best battle plan never survives first
contact with the ... existing code.
This message summarizes our latest triumphs as a precursor to the next
big step, namely submitting patches to begin incorporating our work
into the GCC project.
Repository:
https://gitlab.cobolworx.com/COBOLworx/gcc-cobol/
Ubuntu/Debian x64 binary package:
https://gitlab.cobolworx.com/COBOLworx/gcc-cobol/-/packages/2
Perhaps our most important accomplishment is better support for
Abbreviated Complex Relation Conditions. I mentioned this challenge at
FOSDEM in February. These have the form
A = B or C
which means
A = B or A = C
The standard form is now solved. A nonstandard form that includes
parentheses is now accepted too, and might also be solved. Time and
code will tell.
We continue to add ISO features adopted in recent standards. gcobol
now supports 01 constants, typedefs, and function prototypes.
gcobol supports an EBCDIC mode, whereby ASCII (or UTF-8, or CP1252 )
source operates on EBCDIC-encoded files. That mode has been tested
successfully with the NIST test suite.
gcobol has a documented but as-yet unexploited ISAM plug-in API to
support ISAM libraries for I/O. This opens the potential for ISAM
files that are on-disk compatible with that of other COBOL compilers.
There are numerous improvements to the Compiler Directing Facility
(CDF). Conditional compilation has seen more testing, and commensurate
improvement. It now supports the COBOL-WORDS directive, which is useful
in compensating for non-ISO COBOL dialects.
Speaking of dialects, several concessions are now made to IBM and
MicroFocus dialects. Examples include IBM's LENGTH OF and TALLY
registers, and $IF and friends for MicroFocus.
Contact with the real world forced us to pay closer attention to
performance. Source modules of 400,000 lines of COBOL input no longer
present a problem. Work is ongoing to improve the size and
compile-time speed of the generated code. The GENERIC we produce is
evidently hard on the optimizer.
Users also demand convenience!
-fsyntax-only does what you might guess. That feature allows the
compiler to proceed with parsing in the face of Data Division errors.
-fdefaultbyte initializes memory to a specified value.
-copyext sets the extension of copybook files
-fcobol-exceptions accepts a list of Exception Conditions
-include prepends COBOL source code to the input
-preprocess runs any filter against the amalgamated input, prior to
parsing.
gcobol includes "gcobc", an emulation script for (some) compatibility
with GnuCOBOL.
High on the TODO list is to take advantage of gcc's error/warning
framework. Perhaps in September. We are ever hopeful.
As indicated, we are preparing to submit patches for inclusion of
gcobol into gcc proper. That work will begin in September, once
everyone is back from vacation ready to put nose to grindstone.
Thank you for your kind attention. May the parse be with you.
--jkl
