[asdf-devel] RMCL vs. ASDF

[Pascal Costanza]

On 23 May 2011, at 00:29, Faré wrote:

> Dear Pascal,
> 
>> Here is what happens in RMCL when the configuration says :inherit-configuration. At startup, I get an error message that says this:
>> 
>> ;  NOTE: *TERMINAL-IO* was #<CCL::INITIALIZATION-OUTPUT-STREAM #x4453B26>, was reset to #<CCL::TERMINAL-IO #x4A0FE26> .
>>> Error in process Startup: File #4P"home:" does not exist.
>>> While executing: TRUENAME
>> 
> Which version of ASDF are you using? Following binghe's and Terje's
> messages, I committed in 2.015.4 something that should hopefully make
> it work on RMCL. Please try the latest (currently 2.015.5) and tell me
> if the problem persists.

I will test this as soon as it percolates into Quicklisp.

>> This is triggered by the function user-homedir. While RMCL is still in debug mode for this error, (truename (user-homedir-pathname)) fails. However, afterwards, it returns correctly (for some value of "correct"):
>> 
>> ? (truename (user-homedir-pathname))
>> #P"Macintosh HD:Users:costanza:lisp:RMCL 5.2.1:"
>> 
>> My guess is that user-homedir-pathname returns #4P"home:" as a constant value, even before the logical pathname translation for it is defined during startup.
>> 
> That's strange, and probably a bug in RMCL that it should work at one
> point and not the other. However, ASDF 2.015.4 and later should be
> avoiding (user-homedir-pathname) altogether. Please try again.
> 
>> The deeper problem is, again, that the notion of a home directory doesn't make sense in RMCL (at least not by default). If ASDF wants to have a notion of a home directory, it should be a configurable parameter of ASDF that in most systems can indeed default to the result of (user-homedir-pathname). But it should be possible to explicitly set it in some way before ASDF decides to go looking for it. (This would make ASDF more robust for potential future CL systems that may not have a notion of a home directory either, however unlikely this may be.)
>> 
> When such a new CL system appears, if ever, we can amend ASDF. It's
> not like ASDF is currently devoid of reader conditionals. I had to
> build quite a few ported abstractions to work around the lack of
> portable alternatives.
> 
> —♯ƒ • François-René ÐVB Rideau •Reflection&Cybernethics• http://fare.tunes.org
> As the Chinese say, 1001 words is worth more than a picture.  — John McCarthy

--
Pascal Costanza