Reply #16: Final Assignment - Compilation in Scheme [View All]

One thing that I must point out is that the compilation and machine
execution environment is very fragile and finicky. I discovered
that, sometimes, the environment is left in a strange state after
encountering bugs, such that, after fixing the bug bad behaviors
continue to occur, even though the bug has been fixed. I spent a
lot of time tracking down a subtle bug, then finally exiting scheme
and re-entering, only to find that the bug was no longer there.

More insidious and dangerous is the opposite behavior. A corrupt
environment can lead to correct behavior, even when bugs still
exist in the code. An example of this was when I discovered that
set-car! was left out of the list of primitive functions. I
modified eceval-support.scm to add it, reloaded machine-shell.scm,
ran it, and it worked! Then, I exited scheme, entered it again,
re-ran the same machine-shell, and got a new complaint that
set-cdr! was missing!!! How did it work before, after fixing only
set-car!? I have no idea.

I am suspecting that a lot of the testing I performed yesterday was
against a corrupt environment, because I was able to run s450.scm
successfully, even when some primitives were missing, and when
let->appl was not working properly. I know that I tested the
(define a 1) yesterday and saw it succeed, but this morning after
reading professor XXXX, retested and saw that it was
definitely broken. It is also possible that an error dropped me out
of the s450 environment and on to regular scheme where, naturally,
everything worked just fine... Who knows...