Porting #3

Notes on porting Jam - revised 12/31/2000

1) Working out system dependencies in the Jam code.

    Jam's OS footprint is fairly small.  For OS independent work Jam
    liberally uses standard libc functions like stdio, malloc, and
    string.  The OS dependent interfaces are:

	From filesys.h:

	    file_parse() - split a file name into dir/base/suffix/member
	    file_build() - build a filename given dir/base/suffix/member
	    file_dirscan() - scan a directory for files
	    file_archscan() - scan an archive for files
	    file_time() - get the timestamp of a file, if not already 
			  done by file_dirscan().

	From execcmd.h:

	    execcmd() - execute a shell script
	    execwait() - wait for any outstanding execcmd()'s.

    The current implementations are:

	    filemac.c - mac MPW 
	    filent.c - NT 
	    fileos2.c - OS/2 
	    fileunix.c - all UNIX
	    filevms.c - VMS

	    execmac.c - mac MPW
	    execunix.c - UNIX, OS/2, NT
	    execvms.c - VMS

2) Defining OSMAJOR, OSMINOR in jam.h

    So that the Jambase and Jamfile know their host, Jam defines $(OS)
    to be something useful for each platform.  Make sure that there is
    code in jam.h to generate a useful value for $(OS), and key it off 
    the platform specific C-preprocessor symbol.   If the C-preprocessor 
    doesn't itself defines such a symbol, add a define to the Makefile.

    In addition to $(OS), you can also set $(OSPLAT) if the OS runs on
    multiple platforms (like Linux or NT).

3) Working out system dependencies in the Jambase

    With the value of $(OS) available, the Jambase can be extended to
    support special variables or rules for new platforms.   See the
    current support for VMS, NT, and Mac.

4) Yacc troubles

    The generated files jamgram.h and jamgram.c are distributed for the 
    poor souls without yacc.

5) Known problematic systems:

    - Pyramid has no malloc.h, memory.h

    - Encore has no stdlib.h

    - Bull DPX has sys/file.h problems

6) Send the results back.

    If you do porting work, the result can be integrated into future
    releases if you send it back to the author's address in the README.