[cmucl-help] 19c alpha [Was Re: CMUCL 18c building on tru64 5.1]

[Fausto Saporito]

2014-09-09 20:01 GMT+02:00 Raymond Toy <toy.raymond at gmail.com>:
>
>

>>
>> So I should update undefineds.h with all those functions present in
>> the warnings, if I understood correctly...
>
>
> Ugh. This might be tricky.  Your 19c linux binary supports the
> :linkage-table *feature*. But alpha doesn't. I'm not sure how well
> cross-compiling from linux to alpha will work in this case. And I don't
> remember what a cross-compile without :linkage-tables looks like anymore,
> and all of the current working ports use :linkage-tables.
>
> So, these undefined symbols might be ok, but they do have to get resolved in
> some way.  call_into_c ought to exist in your lisp.nm file. Perhaps genesis
> is not parsing that file correctly?
>
> (See load-foreign-symbol-table in compiler/new-genesis.lisp.)
>>
>>

Sorry, but I'm a bit lost.
Just to recap.

On Linux machine:

1) in alpha-target directory I have lisp.map and lisp.nm Lisp.nm is
empty just a line saying "Map file for lisp version 0".
Lisp.map is full of informations. I noticed lisp.nm is populated later
on the ALPHA machine.

Is this correct ?

When the load-foreign-symbol-table should read that file ? If it tries
on the Linux server the file is empty... and I don't think it reads
the file on the ALPHA machine, because the first step I do is
rebuild-lisp.sh (so only the C code).

On the ALPHA machine:

1) after the rebuild-lisp.sh, I have a lisp.nm with all the infos
needed. Maybe I should put back that file on the linux server and
generate a new core ?

regards,
Fausto