Skip to content

Latest commit

 

History

History
52 lines (41 loc) · 8.22 KB

README.md

File metadata and controls

52 lines (41 loc) · 8.22 KB
Raw

Pascal8000

Conversion of the mainframe IBM 370-based Pascal compiler to run under Free Pascal

It also includes the mainframe-related components, and is an effort at restoring a piece of mainframe software history. Call it a minor software archaeology

There are two flavors of this compiler, the Linkage Editor version, and the Load-and-go version. The load-and-go version compiles a Pascal program then immediately executes it. The Linkage Editor edition acts more like a typical compiler of that time, it creates a load module, which can call external programs if needed, then link edited to form an executable. I had decided to focus on the latter, a compiler that makes actual modules that can be link edited and run independently. My objective in bringing it to the PC was to make as few changes as possible, to keep the compiler "clean enough" that the one I modified can recompile itself, at least, the original, unmodified source. The code it produces is for the IBM 370, not the PC. (It's more a "proof of concept" than anuthing routinely usable. If you have the Hercules mainframe emulator with MVS 3.8J and TK5, or the Z390 Mainframe Studio, you could run it there. (Actually, there already is a copy of Pascal 8000 on TK5; I have some JCL here that could be copied to it and used there to run programs on that machine. See the item on the file Pascal8000B.het listed in the files section of this document.)

My interest in this compiler is more than academic; I am interested in two of its features, ones that were common on older compilers but tend to be missing now.

  • First, the compiler creates a listing of the program, (including a block analysis).
  • Second, when it detects an error, it reports it, then, instead of quitting, skips over part of the code to find a spot to recover, and resume scanning for more syntax or usage errors.
  • Accepting either (* *) or { } for comment blocks.

I'd like to be able to extract and resurrect these features. Then take these from the XDPW / XD Pascal compiler to add more advanced features, like

  • Separate compilation, moving code into other units.
  • Being able to reference published constants, types, and variables from other units.
  • Functions are now "first class" and a function can be referenced indirectly, so that any function with a real number as an argument and returns an integer, can be replaced by any other function with th same type of argument and result.
  • Every function contains a local variable named Result, that has the same type as the function, and is an alias for the function's return value.
  • Strings and concatenation, at least at the 255-byte length.
  • // as a line comment

There are a few things I like from Bernd Oppolzer's New Stanford Pascal Compiler

  • C-style /* */ comments.
  • Strings of up to 32 or 64K, I'm not sure which. He uses a "double length" indicator before each string, two halfwords, one indicating the current size of the string, and the other, the maximum size it has ever been. Thus a string which has been shrunk or nulled can be refilled without having to request more memory.

And from Mad Pascal:

  • Conditional compilation {$IF, {$ELSE, {$ENDIF, and ($ compile-time variables.
  • Include files for repeated code.

A few others I've noticed:

  • More intrinsic functions, ones where a procedure or function call is replaced by the instruction it represents, e.g. inc(x) translated into effectively ++x, i.e. X := X+1.
  • When using a variable or constant in another unit, be able to explicitly reference the item by its unit name, i.e. X in unit There, can also be referenced in another unit as "X" (if there's no other local variable of the same name, either inside a procedure or function where the reference is made, or at Unit level, or a intermediate unit compiled after there), or as "There.X"
  • Given the fact it is a compiler, it might be a good idea to split it into separate pieces, like making scanner and parser separate from code generation; initialization (for the scanner and parser) and local initializaton (for code generation), if it was intended to be able to be modified to create code for some other processor / operating system.
  • The preceding is probably more work than I'd want to do with this. Converting its usages to be compilable by Free Pascal was a good idea; it ensures the origginal code as written would correctly compile. Beyond that I'd probably just want the pieces that work to use them where appropriate in a modern-day compiler with current extensions, with compile-time flags to be sure it retains strict compliance with ISO 7185 Standard Pascal, or where it doesn't (it might have been created before the standard was finalized), bring it up to the standard. But what has been done now will remain in this format, for the language it compiles and the machine it compiled for.

Files .

  • As Pascal8000 has two versions, there are two different compilers, Pascal8000LinkEdit.pas for the Linkage Editor-capable version, and Pascal8000LoadGo.pas for the compile-and-go version. There are also separate run-time systems and initialization for each. It is kind of amazing to think that they developed a compiler that can read and scan a program, create a binary objevt file (or set up the environment to run the code it created) and produce a listing, all in less than 7000 lines of source code.
  • The file $pasmsgs.txt is the error message file used by the compiler. It is the original text, in all upper case.
  • The file pasmlow.txt is the error message file in modified 'Sentence Case.' All lines are lower case except first letter, and any keywords and compiler switches.
  • The file AAEC B1107.pdf is the October, 1979 montly newsletter for a European computer center. Page 13 has an article about Pascal 8000.
  • The file Pascal8000.pdf is the reference manual for the compiler.
  • The directory rsource contains the assembly language support files for Pascal8000 programs. That is not a misspelling, the original PDS (mainframe equivalent of a subdirectory) was Pascal.rsource. The *.asm files are the assembly language sources for run-time support, and the *.mac files are macros they use.
  • The group of files named AAEC_COMPILER are my attempts to make the Linkage Editor Version successfully compile under Free Pascal. At this time it does not work because I stopped to create this release. Some things on it are temporary placeholders for things that actually do things, like opening files. Date and time are properly collected. Also, every character string used in the construction of object files must be converted to EBCDIC, if it was to produce equivalent output.
  • The *.jcl files are the original OS/VS1 and MVS job control language files to extract to a mainframe computer the files from the tape of the compiler, and how to install it.
  • The file Pascal8000B.het is the tape contaning the entire system, and all sources. It can be read by MVS TK5 on Hercules. This is what is used by the previous item.
  • The file Linkedit.obj is the binary object file of the compilation of the linkage editor version of the compiler. It can be link edited with the run-time library to rebuild that version of the compiler. This is the exact onject file generated by this compiler.
  • The file Loadgo.obj is the binary object file of the the compilation of the load-and-go version of the compiler. It can be link edited with the run-time library to rebuild that version of the compiler. This is the exact onject file generated by the linkage editor version of the compiler.
  • The files named bytelocation were a test to see how integer values are stored in little endian computers, like the PC, vs. how they appear in big endian machines, like the IBM 360/370/390/zSystem mainframes. Apparently, 4-byte integers are stored in order 1, 2, 3, 4 in big endian hardware, and 4, 3, 2, 1 in little endian hardware.

Ownership

While the code and documentation contain copyright notices, the Australian Atomic Energy Commission no longer exists, and I am fairly confident that if there is a successor organization, they're not interested in a 46-year-old piece of software having essentially zero commercial value. It is being used as if it was either in the Public Domain or under a BSD/MIT license with no requirement for attribution. I am also waiving copyright on anything I have added as well.