Re: [clisp-list] ffi support on linux mips64 / mips32

SourceForge Headquarters 225 Broadway Suite 1600 San Diego, CA 92101 +1 (858) 454-5900

-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1

Hi Lui,

Lui Fungsin wrote:
> I've sent two emails to cli...@li... and none of
> them show up on the list.

I see at least one
(<http://permalink.gmane.org/gmane.lisp.clisp.general/12060>).

> I'm not sure why but on clisp's homepage
> there's no mentioning of how to subscribe to the mailing list (I
> assume it doesn't require subscription to send a mail to the list.

we cannot have _everything_ on the main page.
it does, however, has the following links:
 -- How do I ask for help?
(http://clisp.cons.org/impnotes/faq.html#faq-help)
 -- mailing list (http://sourceforge.net/mail/?group_id=1355)

both eventually lead to the subscription pages.

> I'm trying to compile clisp on mips64 linux 2.6 and I got this mmap error
>
> Cannot map memory to address 0x4000000000000 .
>
> Searching google I found that other people have similar issues with
> amd64, but it reports that it is fixed. I know that mips64 + linux is
> not a popular platform so probably this is still a issue.

please see clisp/unix/PLATFORMS, "Hints for porting to new platforms" at
the end.
specifically, you may want to start with
$ ./configure --disable-mmap

> warning: memory size out of range, using 4294967295 instead
> Cannot map memory to address 0x4000000000000 .
> [spvw_mmap.d:359] errno = ENOMEM: Not enough memory.
> ./lisp.run: Not enough memory for Lisp.
> make: *** [interpreted.mem] Error 1

look at src/lispbibl.d and adjust your memory settings, specifically
oint_type_shift
oint_type_len
oint_type_mask
oint_addr_shift
oint_addr_len
oint_addr_mask
oint_data_shift
oint_data_len
oint_data_mask

> #./configure
> ......
> (cd .libs && rm -f libcallback.la && ln -s ../libcallback.la
libcallback.la)
> gcc -g -O2 -I. -I../../ffcall/avcall -c ../../ffcall/avcall/minitests.c
> In file included from ../../ffcall/avcall/tests.c:24,
>                 from ../../ffcall/avcall/minitests.c:2:
> ./avcall.h:341: error: duplicate member 'anum'
> In file included from ../../ffcall/avcall/tests.c:24,
>                 from ../../ffcall/avcall/minitests.c:2:
> ./avcall.h:420:1: warning: "__av_start1" redefined
> ./avcall.h:415:1: warning: this is the location of the previous definition
> ./avcall.h:581:1: warning: "__av_reg_struct_return" redefined
> ./avcall.h:574:1: warning: this is the location of the previous definition
> ./avcall.h:589:1: warning: "__av_start_struct3" redefined
> ./avcall.h:577:1: warning: this is the location of the previous definition
> ./avcall.h:846:1: warning: "av_float" redefined
> ./avcall.h:815:1: warning: this is the location of the previous definition
> ./avcall.h:856:1: warning: "av_double" redefined
> ./avcall.h:826:1: warning: this is the location of the previous definition
> ./avcall.h:1172:1: warning: "__av_struct_leftadjusted" redefined
> ./avcall.h:1137:1: warning: this is the location of the previous
definition
> ./avcall.h:1182:1: warning: "__av_struct_rightadjusted" redefined
> ./avcall.h:1145:1: warning: this is the location of the previous
definition
> ./avcall.h:1192:1: warning: "__av_struct" redefined
> ./avcall.h:1153:1: warning: this is the location of the previous
definition
> make: *** [minitests.o] Error 1
> Configure findings:
>  FFI:        no (user requested: default)
>  readline:   no (user requested: default)
>  libsigsegv: no, consider installing GNU libsigsegv
>
>
> So this is the duplicate definition in the .h file
>
> #if defined(__mips__) && !defined(__mipsn32__)
> #define __av_start1(LIST)                                               \
>   (LIST).anum = 0,                                                     \
>   (LIST).aptr = &(LIST).args[0],
> #endif
> #if defined(__mipsn32__) || defined(__mips64__)
> #define __av_start1(LIST)                                               \
>   (LIST).anum = 0,                                                     \
>   (LIST).farg_mask = 0,                                                \
>   (LIST).darg_mask = 0,                                                \
>   (LIST).aptr = &(LIST).args[0],
> #endif
>
>>From the comments in various places it seems that clisp assume the
> platform is SGI if the process is mips.

actually, it appears that at least dec ultrix on mips is supported.

> Anyway, I deleted
>
> #if defined(__mips__) && !defined(__mipsn32__)
> #define __av_start1(LIST)                                               \
>   (LIST).anum = 0,                                                     \
>   (LIST).aptr = &(LIST).args[0],
> #endif
>
> and I can get further until I hit another bummer in avcall.h, which is
> fixed in a similar way by deleting the duplicated definition
>
> So I go into ffcall/avcall/ and run configure, make, make check
>
> ./minitests > minitests.out
> LC_ALL=C uniq -u < minitests.out >
minitests.output.mips64-unknown-linux-gnu
> test '!' -s minitests.output.mips64-unknown-linux-gnu
> make: *** [check] Error 1
>
> It seems that it's checking against some predefined results and that
failed.

normally the minitests output should be empty, but it is not for you.
what is it?

Thanks.
Sam
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.4.6 (GNU/Linux)
Comment: Using GnuPG with Mozilla - http://enigmail.mozdev.org

iD4DBQFHYrLdPp1Qsf2qnMcRArXRAJ98LuIsFqqzYg5zQiWVnRculbVS1wCYmeoc
VzNvGOeqmRWiJ2/mpZQ/mg==
=ovae
-----END PGP SIGNATURE-----