Re: [Axiom-mail] Re: `)read' and large inputs

[Vanuxem Grégory]

Le jeudi 16 novembre 2006 à 13:49 +0100, Ludovic Courtès a écrit :
> Hi,
> 
> Themos Tsikas <address@hidden> writes:
> 
> > The Library domain is for putting AXIOM values from different domains into 
> > one 
> > Keyed-Acess-File. It's a useful way for transmitting values from one AXIOM 
> > session to another. Your first session should have created a directory 
> > called 
> > data.KAF containing a file called index.KAF and in this file would be a 
> > Lisp-readable representation of the values you put in it. Do you get these 
> > files at the end of the first session?
> > The second session must be started at the same working directory as the 
> > first. 
> > Is there a directory called data.KAF when you start the second session?
> 
> Yes, but it only contains a 23-octet long `index.KAF' containing the
> following line:
> 
>   20                  NIL
> 
> The issue is that the first session (where `axiom-large.input' is passed
> on axiom's standard input) _does_ create the `data.KAF' directory and
> the `index.KAF' above but it outputs the following error message:
> 
>    >> Error detected within library code:
>    File is not readable
>    "data.KAF"
> 
> This message does not correspond to any actual read/open error from what
> I can see from the output of `strace'.

This is a bug. The problem is that you are trying to work with KAF files
which involve directories as explained by Themos Tsikas. So all function
such as writable? readable? are applied on this directory and not on the
index.KAF file. The behavior of GCL in this regard has changed (since
2.6.7 I think) : it's not possible, actually, to (open ...) a directory
(readable?, internally, uses this) nor to use (probe-file ...) on it.

I'm CC'ing this mail since Waldek Hebisch (will) work on issues related
to this problem and I don't know if it is aware of this problem.

I _think_ that even if you try to modify the Spad file to avoid the
readability/writability tests your library file (KAF) will be unreadable
because of another problem to open (for reading) these types of file.
This bug was fixed by Waldek but actually this fix is only available in
the svn wh-sandbox branch which is highly experimental.

Sorry,

Greg

PS: I think too (I can not test, my version of Axiom has the same bug)
that some objects can not be saved and retrieved, for example
continuedFraction(0). Themos, am I correct?