Re: [Chicken-users] Solaris 8 tcp.scm matter, and CCAN

[felix]

William Annis wrote:
>         I should mention that I've had some problems with solaris-gcc
> 3.2.1 compiling chicken.  I had to revert to an older gcc to compile.
> With 3.2.1, it barfs out here:
>
> ----------------------------------------------------------------------
> /bin/bash ./libtool --mode=link gcc  -O3 -fomit-frame-pointer -DHAVE_ALLOCA_H 
> -fstrict-aliasing    -o libchicken.la -rpath /u/annis/local/lib 
> -version-info=0:0:0 -no-undefined runtime.lo library.lo eval.lo syntax-case.lo 
> modules.lo profiler.lo  -lm -ldl -lm
> rm -fr .libs/libchicken.la .libs/libchicken.* .libs/libchicken.*
> /usr/ccs/bin/ld -G -z defs -h libchicken.so.0 -o .libs/libchicken.so.0.0.0  runtime.lo library.lo eval.lo syntax-case.lo modules.lo profiler.lo  -lm -ldl -lm -lc 
> Undefined                       first referenced
>  symbol                             in file
> C_toplevel                          runtime.lo
> ld: fatal: Symbol referencing errors. No output written to 
> .libs/libchicken.so.0.0.0
> make[1]: *** [libchicken.la] Error 1
> make[1]: Leaving directory `/u/u01/a/annis/local/src/chicken-0.1082'
> make: *** [all] Error 2
> ----------------------------------------------------------------------
>
> When I use gcc2.95.2, everything compiles fine.  I've not yet had any
> time to dig into this curiosity.

Urgh. Well, the reorganization of the library stuff (which Jonah is currently
doing) might definitely help here.

Sorry, I haven't got an answer to that, yet.

>  >It already works (as far as I have tested it). I've just cleaned up
>  >the output of the setup-code and I plan to put a little selection of
>  >extensions on call/cc.org, Real Soon Now.
>
>         Cooler yet.  I'll be creating some libraries for Unixy sorts
> of things as I work with chicken more.  A few network protocol
> libraries would be handy, too, such as basic FTP and SMTP tools, since
> I regularly write programs that send me email or shuffle files about. 

Wow, that would be cool! I have only a few eggs ready yet, but enough
to start a collection, anyway ;-):

regex-case       - the regex-case macro that used to be in Chicken
define-datatype  - from the EOPL book
tcp              - them sockets
base64           - base64 strings
http             - a *very* basic HTTP client/server library
awk + make       - both stolen from PLT, in turn stolen from Scsh

I will make them available, as soon as the new release is out, and
when I have prepared some HTML for an "Egg" repository.

Call for assistance:

"Egg's Unlimited" might be a worthy title. Or is it too silly?

And: can someone provide an image of an egg with a lambda?

;-)

[Hm, I think I'm overstressing the "egg" thing a little...]


>         Although this makes me wonder about how to approach these.
> OOP?  Something else?  I detest using OO for the sake of OO, although
> that seems to mesh reasonably well with this sort of programming.  At
> any rate, a consistent approach seems best.
>
>         So, what to people think?  Network protocols via OOP, or via
> functions called on handles?


If you intend to provide a collection of related extensions that might
share a common interface, OOP might be a good option. Personally I
prefer handler procedures, but mostly out of rather subjective
"efficiency" concerns (which not necessarily reflect real
performance ;-). TinyCLOS (if you intend to use that) is not
exactly superfast, but *very* flexible, and if performance is not
of utmost importance, I think it is a viable alternative.


cheers,
felix