[Sbcl-devel] Re: PATCH: Support for SBCL HEAD (and future 0.9.2)

[Nikodemus S. <nik...@ra...>]

On Thu, 9 Jun 2005, Juho Snellman wrote:

> Unfortunately it also breaks compiler note annotations when doing C-C C-C.

"oops"

>  * Since the source is being LOADed from a string-input-stream,
>    LOCATE-COMPILER-NOTE in swank-sbcl needs to handle the case where the
>    compiler error context doesn't have a filename. Maybe something like:
>
>    (cond ((or (and (pathnamep file) *buffer-name*)
>               #+swank-backend::source-plist
>               (eq *buffer-name* (getf sb-c::*source-plist* :emacs-buffer)))
>           ...)
>          (...))
>
>  * Now that we do a LOAD instead of some form of COMPILE, source path
>    information is no longer preserved. LOAD calls EVAL on each form,
>    and EVAL does a macroexpand before the form is passed on to the
>    compiler.

FWIW, I think these are both more of an SBCL issue then Slime issue: while 
it may indeed be reasonable for SWANK-COMPILE-STRING to move back to

  (COMPILE NIL `(LAMBDA () ,@(READ-FROM-STRING ...)))

or equivalent, I'll have a look at getting LOAD & EVAL play nice(r) with 
source locations.

Cheers,

  -- Nikodemus              Schemer: "Buddha is small, clean, and serious."
                   Lispnik: "Buddha is big, has hairy armpits, and laughs."