RE: [Axiom-developer] boottocl

[Bill Page]

> | On September 30, 2006 10:36 PM Gaby wrote:
> | > ... 
> | > Assume I have an issue with how boottocl works, which version
> | > am I supposed to modify?  I'm talking of the system as it
> | > currently is -- not how it could theoretically be.
> | > 
> | 
> "Bill Page" <address@hidden> writes:
> | If you were planning to change anything significant in Boot
> | I would suggest that you plan on *first* modifying the build
> | so that the bootsys step with the old compiler is not longer
> | necessary, i.e. start with the new Boot in interp. Then there
> | will only be one  version to maintain and the build process
> | will be (slightly) faster.
> 

On October 1, 2006 9:43 PM Tim Daly wrote:
> 
> in boot/ptyout.boot there is a function in the package BOOTTRAN
> called BOOTTOCL.
> 
> ptyout/bootran::boottocl
>  -> boottocllines
>   -> shoeClLines
> ...
> in util.lisp [interp/util.lisp]there is a function in the
> package BOOTRAN called BOOTTOCL.
> 
> util/bootran::boottocl
>  -> boot::boot
>   -> bootlex/boot::PARSE-Expression
>    -> new2oldlisp
> ... 
> first, bill burge loved puns and i remember him explaining to
> me that the latest version of boot was much better and more
> comfortable, thus he was changing over to calling it 'shoe'. this
> 'shoe' shows up in the src/boot/ptyout version.
> ...
> thus it seems there are 2 boot languages implemented (but not 
> defined). my belief (but i don't remember) is that the src/boot
> directory came into being as part of the process of switching to
> the new boot/shoe language.
> 
> you'll notice that the makefile in the src/boot directory explicitly
> translates using the BOOTSYS image which contains the 'shoe' parser.
> the makefile in the src/interp directory explicitly translates using
> the DEPSYS image which only contains interpreted lisp files from
> the interp directory and is NOT the same as the boot/shoe version.
> 
> rather than referring to them both as boot perhaps we should
> call the one in the src/boot directory 'shoe', following bill's
> convention.
> 

Thank you Tim!

On October 2, 2006 1:38 AM Gaby wrote:
> Based on what Tim said, I don't think I'm going to gain much
> from eliminating "old" boot at the moment. The build time, as
> I see it, is dominated by building the algebras.  And each algebra
> file seems to me to take forever to compile.  Making compilation
> faster is going to speed up the built process.
> 

Let's us Tim's suggested terminology. "Shoe" is the newer version
of Boot in the boot directory, while "Boot" is the older version
of Boot in the interp directory.

My suggestion was to eliminate the source in boot directory
(thus remove Shoe) and use the same method as in the CMUCL build
starting from Boot in the interp directory. Thanks to Tim's analysis
we now know that this would amount to removing the incomplete attempt
to move to a newer version of the BOOT language implied in Shoe.

I would still argue that this is a good idea. It has nothing to
do with speed of the build process. I think the current build
process for Axiom is too complex and eliminating one step results
in a significant reduction in this complexity. It also eliminates
some uncertainty because we no longer have anyone with any
expertise in what Bill Burge was trying to accomplish with Shoe.
The Boot compiler that is in interp is (currently) an integral
part of the SPAD compiler. It would seem to me that eliminating
it would be much more difficult and probably the reason why the
introduction of Shoe was never completed.

Regards,
Bill Page.