[asdf-devel] new asdf does not like matlisp.asd

[james anderson]

On 2010-04-11, at 21:57 , Robert Goldman wrote:

> On 4/11/10 Apr 11 -2:51 PM, Mario S. Mommer wrote:
>>
>> Hi,
>>
>> The error with the fd-stream is ignorable. In slime just let the  
>> buffer
>> with C-x 1 while in the repl, and that's it. As far as I know,  
>> that is a
>> gfortran bug that is not there in the next version. If you bury the
>> buffer with that error you can continue to work without
>> trouble. Strange, but true. I've been doing that for about a year :-)
>>
>> If that does not work, I'll attempt to do a lisp-only or some other
>> non-fd-crashing thing.
>>
>> (re aclrepl: maybe you built sbcl with sb-aclrepl?)
>
> There seems to be a lot of contingent stuff here that may be obscuring
> any ASDF issues.
>
> Question: is it possible to get an analogous issue using only lisp  
> files?

not in the standard configuration.
the unix-dso has output files, which result causes operation-done-p  
to assert that it is never done.
a vanilla module does not.

however, take the matlispbug system definition,
- strip out everything except the file, components, and module terms,
- declare all the files to be :file, declare all the modules to  
be :module,
- make the proper arrangements in the file system,
_and_ add a gratuitous in-line output-files method to the initial  
modules to return a list of files,

and you will observe that an operation gets carried through  
dependency on the module.