Re: Building with gcl recompiles everything?

On 4/10/26 3:25 PM, Raymond Toy wrote:

> On 4/10/26 1:55 PM, Robert Dodier wrote:
>
>> On Fri, Apr 10, 2026 at 1:24 PM Raymond Toy<[email protected]> wrote:
>>
>>> I commented out the rule for tools/sys-proclaim.lisp, and then touched cl-info.lisp. Building now only compiles cl-info.lisp and quickly loads an the remaining fasls. Very fast.
>>>
>>> Perhaps that’s the solution for now. IIRC, the sys-proclaim stuff just helped gcl generate better code by declaring the correct function arg and return types.
>> I gather that sys-proclaim is not in the picture in this most recent
>> build. Given that the purpose of it was to help GCL generate more
>> efficient code, I wonder how much difference it makes. Can you try
>> running run_testsuite with the new and previous builds? I am aware
>> that run_testsuite is not a comprehensive speed test but anyway it is
>> what we have available at the moment.
>
> This will take a while.
>
Still building…. (My old laptop isn’t all that fast anymore.)
>
> I do wonder if sys-proclaim is working. When I build maxima, I get 
> lots of compiler notes that the function signature has changed. It’s 
> been that way for some time now.
>
Well, it doesn’t. Take a look at defun emit-fn 
<https://cgit.git.savannah.gnu.org/cgit/gcl.git/tree/gcl/cmpnew/gcl_collectfn.lsp#n69>. 
This function is supposed to cause the compiler to emit .fn files. I 
can’t find them anywhere because that function does nothing. And git 
annotate says that was changed in commit 5d9220be12 in 2021-04-15.

I guess this means it hasn’t been working for 5 years, and no one noticed.

I propose we get rid of sys-proclaim completely. If Camm wants to make 
it work again, I’ll let him make the changes.

&#8203;

_______________________________________________
Maxima-discuss mailing list
[email protected]
https://lists.sourceforge.net/lists/listinfo/maxima-discuss