World: r3wp

the only time i trap about that bug is when i do a script i did before, 
thus overwriting the loading function.

and in that case i prefer to care mayelf isntead of forcing every 
tool to be aware of it.

you don't overwrite the function at all - Rebol functions are anonymous 
in nature, it is just an interpreter bug

ok, overwrite the only reference from the global context.

if a safe version is no effort or even shorter, thats ok, will not 
stop that.

did you notice why the former version above may become a victim of 
the GC bug while not looking like overwriting the reference?

hmm, started re-reading just a moment before. thought maybe i miss 
something.

No, i dont see how. everything is referenced from locals. as long 
as you don't overwrite 'default itself, it should work?

actually not, the problem is this:

default [1 / 0] [default [2 / 0] [...] print mold disarm error]

but i missed the point anyway, thought 'default takes care if i overwrite 
my own function. to hasty. its because of its own created function 
i guess.

oh, recursion. and locals on the recursion-stack are not checked 
too?

aha, it *may* work, I didn't check it

Hope with your bug-example Carl has an easy (and thus quick) time 
to fix it :)

when we use DO method, we are *declaring* the function as anonymous 
and the interpreter takes care, the only problem may occur, when 
the interpreter handles the function as *named*, although that shouldn't 
matter, because every Rebol function is essentially anonymous

it matters because the gc does not scan the c-stack correctly i guess. 
and its using c-stack for nesting.

so the anonymous reference there does not count, is overlooked. You 
need another one, usually that from the functions name. kill that, 
next gc bang. 'do takes more care somehow.

actually DO names the function as its argument in a sense, that is 
why the original name isn't needed

hmm, native functions could have argument-lists like normal ones, 
and then its "named" in that list? could be.

funny, isn't it?

:-)

I believe, since you can expect the crash to be fixed some time in 
the future, that the solution is to provide both versions of the 
function, the fast, vulnerable version commented, and a note explaining 
when to switch from the slow, safe version to the fast, vulnerable 
version (preferably with test code to prove if your interpreter has 
the crash or not).

To me the "sfae but slow"-version with 'do looks more elegant. i 
doubt it is slower. but i may compare the wrong sources, lots of 
them..

Topic - needs or includes.... I am just cleaning up some scripts 
and want to use the dynamic load capability that I thought was enabled 
by pre-requisite setting in the REBOL header.    The only link I 
can find is to the Values.R in the library.   Is there a standard 
approach to this?   Now I am using Carl's %include.r from the script 
library but it does not check for prior loading.

I don't think, it's initially possible to check, if a certain script 
has been loaded or not. One approach, that is often seen in C includes, 
is to have a big 'if' in the script around all the code there, which 
checks on a define, and inside the 'if' define the thing being checked 
on.


I'm searching for a good 'include' myself from time to time. One 
where it's possible to include something without full path. Maybe 
variables like system/options/path, system/options/home or system/options/boot 
should be used.

wrong, my include does it

:-)

Ladislav, clever approach with the included block! Thanks!

check the latest INCLUDE now, http://www.fm.vslib.cz/~ladislav/rebol/include.r

How do you include "include.r" in your projects? Something like putting 
it in the home directory and type:
do join system/options/home %include.r

It should be part of REBOL.

yes, I put it in %user.r together with a definition of a suitable 
INCLUDE-PATH

OTOH, if you mean how can I build a script containing INCLUDE, it 
can be done too, the easiest way is to use #include-string for that 
to prevent the "execution" of the preprocessor instructions defined 
in %include.r

The new Core 2.6.0 in View 1.3 is running the user.r found by the 
environment variable APPDATA, it seems. On my Windows computer, it's 
in C:\Documents and Settings\John\Application Data\rebol\ , even 
if I install REBOL/View in E:\rebol\view\. Is it a good way of doing 
it? I find it a bit confusing.

WinXP, that is. I wonder, how it is under Win98, Win2000 and Linux?

Thank you Ladislav.  More than I had hoped for.   p.s. is there much 
of a penalty for using include/check instead of include?

I didn't measure that but it doesn't matter to me, because Include 
isn't used often anyway

I think, that it is unnoticeable

the slowest part of INCLUDE work is read anyway

Great then I will use include/check as the default usage and may 
save some slow file access time in sub-modules.  Thanks again.

Geomol: you should be able to change the location of user.r and so 
on during installation (note that installation location and user 
files location are two different places, but can be set to the same 
dir)

Geomol: This way the executable may reside on a readonly file system. 
Sounds good to me.

Geomol, that way of locating user.r on Windows is really for the 
best. Windows is a multiuser OS after all, and the APPDATA directory 
on Windows is used roughly the same as the home directory on Linux. 
Global settings can be loaded from rebol.r in the same directory 
as the View executable.

I see, thanks!

Geomol, I have an include system as well.  It doesn't check for already 
loaded scripts but you may want to look at it.
http://www.lexicon.net/antonr/rebol/library/include.r

Can't find www.lexicon.net, it seems.

URL is correct, perhaps try again later.

I would like once again, if find/match on blocks has correct behavior? 

blk: ["Petr Krenzelok" "Richard Smolak" "Ladislav Mecir"]
find/match blk "Richar"


... will not work, and I would expect it to. At least RebDB gives 
me such functionality and it is nice to have ...

doesn't find/match only work on a per element basis? otherwise it 
would be much more complex... if not strings, how would you search 
words, objects, integers, etc.?

if it would work on per-element basis, it would work ...

actually I want someone to explain me, how is supposed find/match 
to work with blocks, if it was ever supposed to work? ;-)