[asdf-devel] pathname test

[james anderson]

good morning;

[ the context for this message is not clear. ]

On 2010-03-16, at 06:33 , Faré wrote:

> This is invalid as a pathname to an actual file, only valid as
> something that you can merge with a pathname that has a type:
> (make-pathname :directory directory :name "typed-file.type" :type nil)

"this" is a file without a type?

>
> Conversely, this is probably not a good idea for naming a file to
> load, as many implementations will add a type of .lisp or .lsp (or
> .fas) before trying to load such a file:
> (make-pathname :directory directory :name "untyped-file" :type nil)

all implementations are permitted to use an incomplete specification  
to designate any extant source and/or binary file and load one of  
them.[0] that passage leaves questions unanswered:
- (stated explicitly) which typed filed to load.
- (unstated) what is the consequence if an untyped file exists as  
well as one or more typed ones.

it does not permit to not load an untyped file if that is the only  
one which exists.

> Or maybe they shouldn't? Not clear from the spec, but common  
> practice is there.

i read nothing in the text which permits to require a type.

>
> In other words, you need a different set of valid values for
> static-file and file components (though there might be a common
> subset).
>

given the definition for load, a source file component should be able  
to designate a file without a type. is there some other distinction?

> Can you work on the test as now part of the official testsuite?

in principle, ok.

---
  [0] : http://www.lispworks.com/documentation/HyperSpec/Body/ 
f_load.htm#load