In 2004 I submitted a CLRFI which was then accepted as CLRFI-2.
It specified a new function COMPILED-FILE-P which was supposed to be
useful for system construction utilities, e.g., asdf:

Function
(compiled-file-p file-name) ==> valid-p

Returns
 true  if the file appears to be a valid compiled file
  (i.e., exists, is readable, and the contents appears to be
  valid for this implementation),
 false  otherwise.

Implementations are required to inspect the contents
(e.g., checking just the pathname type is not sufficient).
Although the completeness of the inspection is not required,
this function should be able to detect,
e.g., file format changes between versions.

I re-submitted the spec to the CDR editors a week ago.
No response so far.

The detailed html spec is attached.

<!DOCTYPE html PUBLIC "-//W3C//DTD XHTML 1.0 Strict//EN"
          "http://www.w3.org/TR/xhtml1/DTD/xhtml1-strict.dtd";>
<html xml:lang="en" xmlns="http://www.w3.org/1999/xhtml";>

<head>
<title>Function COMPILED-FILE-P</title>
<meta name="author" value="Sam Steingold"/>
</head>
<body>

<h1>Function <code>COMPILED-FILE-P</code></h1>

<h2>Author</h2>

<p>Sam Steingold</p>

<h2>Related</h2>

<p>ANSI Common Lisp standard function <code>compile-file</code>.</p>


<h2>Abstract</h2>

<p>A facility to determine whether a file is a valid compiled file for
 the specific implementation.</p>


<h2>Rationale</h2>

<p>Build tools, like <code>defsystem</code> or <code>asdf</code>,
have to determine whether a file needs to be recompiled.</p>
<p>Obviously, when the compiled file is older than the
source file, recompilation is in order.</p>
<p>Alas, there are other situations when this might be necessary, e.g.,
when the implementation changes the compiled file format or when two
implementations use the same name for their compiled files
(<code>.fasl</code> is used by both <code>SBCL</code> and <code>ACL</code>).
</p>


<h3>Current Practice</h3>

<p>Implementation-dependent.</p>

<h3>Cost of adoption</h3>

<p>Probably tiny: an implementation must be able to check for compiled
 file validity, so all it takes is to export the necessary
 functionality, e.g.:</p>

<pre id="compiled-file-p-clisp">
#+clisp
(defun compiled-file-p (file-name)
  (with-open-file (in file-name :direction :input :if-does-not-exist nil)
    (and in (char= #\( (peek-char nil in))
         (let ((form (ignore-errors (read in nil nil))))
           (and (consp form)
                (eq (car form) 'SYSTEM::VERSION)
                (null (nth-value 1 (ignore-errors (eval form)))))))))
</pre>

<h3>Cost of non-adoption</h3>

<p>Users will suffer random errors when trying to load invalid binary
 files.</p>


<h2>Specification</h2>

<p>Function</p><pre>
(compiled-file-p file-name) ==&gt; valid-p
</pre>

<p>Returns</p><dl>
 <dt><code>true</code></dt><dd>if the file appears to be a valid compiled file
  (i.e., exists, is readable, and the contents appears to be
  valid for this implementation),</dd>
 <dt><code>false</code></dt><dd>otherwise.</dd></dl>

<p>Implementations are required to inspect the contents
(e.g., checking just the pathname type is not sufficient).
Although the completeness of the inspection is not required,
this function should be able to detect,
e.g., file format changes between versions.</p>

<h3>Exceptional situations</h3> <ul>
<li>signals an error of type <code>type-error</code>
when the argument is not a <em>pathname designator</em>.</li>
</ul>

<h3>Examples</h3>

<pre>
(compiled-file-p "foo.lisp") ==&gt; NIL
(compiled-file-p (compile-file "foo.lisp")) ==&gt; T
</pre>


<h2>Reference Implementation</h2>

<p>See <a href="#compiled-file-p-clisp">above</a>.</p>


<h2>History</h2>

<p>This used to be <strong>CLRFI-2</strong> (in 2004).</p>

<h2>Notes</h2>

<p>The trivial implementation:</p>
<pre>
(defun compiled-file-p (file-name)
  (not (nth-value 1 (ignore-errors (load file-name)))))
</pre>
<p>is wrong because,</p>
<ol>
<li><code>load</code> may fail even though the file is valid:
   even when <code>foo.lisp</code> contains calls to <code>error</code>,<pre>
(compiled-file-p (compile-file "foo.lisp"))
</pre>should still return <code>T</code>.</li>
<li>this is not side-effect-free, i.e., this may define new functions and
macros (or, worse yet, redefine some existing functions and macros or
execute some malicious code).</li></ol>
</body>
</html>

-- 
Sam Steingold (http://sds.podval.org/) on Ubuntu 12.04 (precise) X 11.0.11103000
http://www.childpsy.net/ http://mideasttruth.com http://ffii.org
http://www.memritv.org http://jihadwatch.org http://camera.org http://memri.org
You cannot fire me. Slaves are not fired. Slaves are sold.
_______________________________________________
pro mailing list
pro@common-lisp.net
http://lists.common-lisp.net/cgi-bin/mailman/listinfo/pro

Reply via email to