[Ecls-list] Latest changes & request for comments

[Gabriel Dos Reis]

On Sun, May 17, 2009 at 6:05 AM, Juan Jose Garcia-Ripoll
<juanjose.garciaripoll at googlemail.com> wrote:

>  - When COMPILE-FILE is invoked with a non-nil value of :OUTPUT-FILE, ECL
>   now honors the file type supplied by the user, instead of overriding it
>   with "fas" or "fasl".
>
>  - When COMPILE-FILE finds that the value of :OUTPUT-FILE has an unsupported
>   file type, it allows the user to register it with LOAD as a valid binary
>   file extension.
>
> Would this be enough to solve the binary file name issue?

I just realized something that may explain our differing views.
It appears to me that you seem to believe that ECL is required to
load the binary file with a user-supplied extension, previously
produced by COMPILE-FILE with non-NIL :OUTPUT-FILE.

I made no such claim in the original bug report.  So, it appears
that the fixes you committed do more work than needed (and
end up not actually working for the cases where it matters).
Here is a concrete proposal:
  (1) keep LOAD logic/infrstructure as is -- no magic numbers and co.
  (2) don't override the file extension if supplied by user.
  (3) if you really feel that ECL should say something about (2), then
       maybe issue a compiler note but don't pause with a restart.

That will solve the file name issue as I reported, and will not require ECL
to change how is determine file loading.

-- Gaby