Re: Porting mit scheme

[Taylor R Campbell]

OK, I finally got around to verifying that the svm1-32le build works
on master.  Had to fix a bug affecting only 32-bit systems that crept
into bytevector.scm -- committed.


> Date: Wed, 6 May 2020 01:04:58 -0400
> From: James Flaherty <kitcat425.314@gmail.com>
> 
> That is entirely my bad. I re-ran the command with the parentheses this
> time. The configure command worked but the make -j4 cross-target command
> returns the following
> (echo '(with-working-directory-pathname "compiler/machines/svm"' && \
>  echo '  (lambda () (load "compile-assembler")))') \
> | 'mit-scheme' --band runtime.com --batch-mode --no-init-file
> /bin/sh: line 2: mit-scheme: command not found
> make: *** [Makefile:356: compiler/machines/svm/svm1-defns.h] Error 127
> 
> it looks as though some part of the makefile is still looking for a
> resident mit-scheme binary to complete the build?

This suggests that you did not ship over the complete state of a tree
in which `make cross-host' has successfully completed, or there is a
bug in the build.  Can you please pull from master, make sure you have
a clean working tree, and share a complete build transcript?

In particular, run the following commands from the top-level directory
of the mit-scheme repository clone, and include their output in the
transcript:

# on the cross-build host
env
git status
git clean -dxn
git -P log -1
cd src
./Setup.sh
./configure --enable-cross-compiling --enable-native-code=svm1-32le
make -j4 cross-host

(make sure this command succeeded; then run `rsync -avzHc' to transfer
the entire mit-scheme tree over, or `tar cf - mit-scheme | ssh ...
"tar xvf -"', or whatever, but make sure the target is clean before
you do that)

# on the cross-build target
cd src
(cd microcode && ./configure)
make -j4 cross-target
env FAST=y make check

You may find it helpful to use the `typescript' command to record the
transcript.