World: r3wp

What did Carl mean about rebcode modifying memory? I'm a little worried 
about these use-an-opcode-wrong, REBOL-goes-away errors.

That  NOT x  crash can be worked around by doing a  SETT x GETT x 
 to make sure the value is compatible.

That or be extra careful with specifying those argument types to 
rebcode functions.

you should always specify argument types for REBCODE functions...

and... use the TYPE? opcode when you cannot determine the type in 
advance.

anyway... i think there's still a lot to improve in this area...

I would prefer opcodes to just don't do anything when given bad arguments, 
instead of crash REBOL with a core dump, as a general rule.

You all have been improving rebcode greatly in that respect so far.

Carl mentioned everything is traded for speed.

Btw speed, little benchmark rebol, rebcode, c: http://polly.rebol.it/test/test/rebcode/download.r

They'll never get REBOL secure if you can cause core dumps with rebcode.

Maybe rebcode is not enabled then, or there are two different interpreters.

Core dumps on Windows are caused by unintentional buffer overflows 
and such. If you can cause them unintentionally, you can cause them 
intentionally.

There are some things you shouldn't trade for speed.

Personally i agree. OTOH lots of other apps do that trade too.

If you want to do multimedia, either trade or have a very smart compiler.

and the rebcode-portions should be a small part of the programm. 
And i guess often generated by dialects. which will add checks on 
their own, which is why we have this range-checks now (IMO)

Well yes, I want to make a very smart compiler as much as the next 
guy, but still I like the invalid-as-noop behavior that many of these 
opcodes have, that all of them should have.

At least invalid-as-error would do, rather than invalid-as-crash.

Heck, in debug mode I would like all of the opcodes to act as invalid-as-error.

hmm.. does IF work?

r: rebcode [] [
  set a -1
  lt a 0
  either [return a][return 0]
] ; works

r: rebcode [] [
  set a -1
  lt a 0
  if [return a]
] ; syntax error

ift, iff (it-true, if-false)

(why not if, unless?)

ah, silly me. knew I shouldn't have read the docs :-)

Maybe they use ift, iff to empharize that they don't work like REBOL 
if, unless, that they operate on condition codes?

But that is true for pick and such true. i mean different args but 
same name. maybe saves typing?

Could be an assembly thing :)

brat and braf are like that

And they came first - the inner block thing came later.

How do I break out of a loop? This won't work:
until [
eq a 0
ift [break]
...
]

Break will exit an 'until:
    until [ break 1 = 4]  ;; does not loop forever.
Have you redefined 'break somehow

this looks like a special case

Sunanda, I'm talking rebcode! Isn't your example non-rebcode?

G: try until [eq a 0 braf 1 break ...]

Whoops!! Sorry!

Ladislav, thanks! That did the job.

rebcode now seems to be regarded an alpha version. Initial release 
contained some fine docs, not included with later experimental releases. 
Will new extended docs come?

Personal tought on REBCODE: "sniiiiiiiiiiiiiiiiiiiiiiiiiiiif don't 
toutch my rebol it looks like ASM sniiiiiiiiiiiif my poor reeeeeeeeeeeeeeeeboooooooooooool 
is all broken  :
                           set i 1
		eq i 1
		braf fail
		brat continue
		print "Should not print!"
		label continue
Sniiiiiiiiiiiiiiiiif "

but performances are great ...

With rebcode we loose on of his key feature the simplicity ... But 
for some tasks really precise that needs high perfs it's great to 
have such a tool . I don't kno yet if i ill use it every day...

you surely shouldn't

I like the brat and barf instructions :-)

but it looks not like rebol anymore  ...

the fun is, that it is REBOL. Have you seen the Alien Dialect?

I still think it's pretty simple, by assembler standards. Have you 
seen x86 assembler?

Urgh

In rebcode, every instruction has just one addressing mode, does 
just one thing. Count your blessings when you can.

Yeah i do a lot of asmx86 but ... I don't like that even if you get 
a close controle on what you are doing

shadwolf - the nice thing is, that for specific domain, you can create 
your own sub-language using parse and let it generate rebcode for 
you :-)

For some kinds of coding I would actually find rebcode to be a more 
comfortable dialect. For instance, you don't have to worry about 
whether someone has redefined add.

shadwolf - really no reason to use it on a daily basis, but cool 
to have for fast tasks - without it real-time image manipulation 
in rebol was impossible for e.g.!