World: r3wp

simplification:

o: make object! [f: does [g print "version 1"]]
g: does [o/f: does [print "version 2"] recycle]
o/f

this looks like being enough too:

f: does [g print "version 1"]
g: does [f: does [print "version 2"] recycle]
f

or shorter:

f: does [g print "version 1"]
g: does [unset 'f recycle]
f

the crash observed in Linux View 1.3 too

...and under OpenBSD

the problem can be "circumvented" as follows:

f: does [g print "version 1"]
g: does [unset 'f recycle]

doer: func [f [function!]] [f]
doer :f

Nice work Ladislav.

Cool reduced. I have that bug sometimes when redoing scriptsbut never 
nailed it really. Never thought to use recyle. my workaround:  f: 
func[][dont-gc-me: :f reload-me]

There is a problem with the issue! type. It doesn't accept the slash 
character as the example in the rebol core manual says.

>> value: #MG82/32-7
== /32-7
>> probe value
#MG82
== #MG82 ;; It droped the 32-7 part.

regarding the above GC crash (and to Volker): I think, that the best 
workaround is

    do :f

instead of just 

    f

Eeks. should this happen?
  echo -e "these are\nsome words" | rebol -q scratch.r
the script contains an error, an undefined word bang
** Script Error: bang has no value
** Near: bang
** Script Error: these has no value
** Near: these are
** Script Error: some has no value
** Near: some words

input is executed. its clear, there is an error, so go to console 
for input. but this time the console is data. maybe with pipes that 
should be changed?

workaround is: run as cgi, rebol -qc . but this is easy to overlook.

if you look in svv/vid-feel, you can see many of the 
OVER functions have arguments like:

	over: func [face action event] ...

but according to old feel doc, they should better be:

	over: func [face action position]

or according to new view system doc:

	over: func [face into position]


This is to avoid misleading error messages such as:   ** Script Error: 
over is missing its event argument

Should I post a report ?

Yes please. These may not be "errors" in and of themselves, but such 
inconsistencies have a ripple effect as you've identified that make 
REBOL harder to use / understand (especially when compounded with 
other inconsistencies). BTW, the other inconsistency I've noted in 
the past in this area is the use of "face action" versus "face act" 
versus "f a" in some places. It would be nice (if not already so) 
if this was sorted out.

Ok, posted it. But on consideration of the issue of abbreviated argument 
names, I don't feel strongly enough that it's so bad. In some cases 
it looks justified to abbreviate FACE to F, like when there are a 
gazillion references to the face. In some cases, the arguments are 
only used once each so there's not much advantage in abbreviating 
them (but they are anyway).

It depends ... if you know enough about Rebol, then "over is missing 
its 'p argument" may be enough. 

On the other hand, if you are just starting, then 'position may help 
you more ...

on linux:
!>load probe mold load "1e-50"
1E-50.0
** Syntax Error: Invalid decimal -- 1E-50.0
** Near: (line 1) 1E-50.0

makes problems with prebol and friends

Doesn't do that on windows -- so definitely a cross-platform problem.

posted it to rambo

;Rebol is gong into limbo with this code. I don't think it should 
happen.
parse {my infinte parse loop} [any [string!]]

The interpreter stops respondin and start consuming a lot of cpu 
cycles.

Don't know about regular apps, but in the console I can interrupt 
that with the Escape key (as of 1.3)

whoa - doesn't let go of the memory, though. I ended up getting a 
Virtual Memory warning from Windows after doing that a few times. 
Memory consumption went down as soon as I quit the Rebol console.

Is there any knowledge when the Problem with open/direct will be 
fixed or what prio this has at RT? I already asked in the Mailinglist 
but got no answer. There was only mentioned that in one Beta it might 
be fixed.  The thing that i always don't understand with Software 
vendors of all kinds is that they always focus more on new features 
than on stability and perfectly working already built in features

RT is re-prioritizing things as they go, because the community will 
get excited about certain things, so it's good go consider putting 
a little time in on those. Other tasks have larger implications and 
take a lot of time to design and code. The /direct issue is a good 
example because they don't want to break file handling code (which 
is used in *many* scripts), nor do they want to just dump in a quick 
change and make matters worse in the long run.

What about a new refinement /stream and keeping /direct for backward-compatibility? 
And the old protocols are source, one could add an option to enable 
them. Would then be a small change to get scripts running again, 
until the bigger one of some rewriting is done.

So, understanding you correctly , there was a time when open/direct 
worked?  I need this to get access to large files without the system 
trying to load it completely into memory/buffering it. Without this 
option there is no way for me to use REBOL for my problem. I don't 
understand where the compatibility problem should be. Open/direct 
is defined to get non buffered access to ports. Was this different 
in the past?  As you say there are scritps where this feature is 
used, so there must be a version where this worked or was it used 
with the wrong behaviour ?. If so, i understand why there is a discussion 
to change the behaviour. But then the definition of open/direct should 
be changed and a new refinement be implemented which gives you the 
possibility to take advantage of non buffered port access. 


I know the community needs new features, not only for marketing purposes, 
but often enough i really don't understand how a priority is set. 
I know, it is not my business to decide such things. I am only a 
person interested in REBOLS future.

Thorsten, I just quickly tested with

 port: open/direct/skip %user.r 3 probe copy/part port 20 close port

and found View beta versions 1.2.54.3.1 thru to 1.2.57.3.1e  do the 
skip correctly. Those versions are the ones with the async core, 
which was found not to be quite stable enough, and it was removed 
for the next beta version, View 1.2.100.3.1

I was just amazed the open/direct didn' t return anything in View 
1.3.1 not even an error. Thanks to this hint. I Think i try one of 
these to keep up development, hoping there will be a fix or new solution 
in one of next releases. Perhaps it will be in the near future ;-)

I hope so too. It's a good bet it will work the same way in future 
versions, when it will be added again.

I would try to get 1.2.57.3.1e, it being the last in that set of 
versions.

the alpha also has /seek.

also, it's not that /direct does not work; it does, always has, in 
that it does not buffer. the problem is that it does not allow seeking, 
i.e. it only allows sequential access. so you can open a 2GB file, 
as long as you read it sequentially.

Where does the 2GB limitationcomes from??

32bit-integers.

Btw, would it help to have a native dll for file-access until rebol 
is smarter? should not be that hard.

That should certainly work Volker. If I had the need under Windows, 
that's what I would do. So far I've been able to live without it.

I am considering adding this wish to RAMBO:

Add EVENT/ENTRY-FACE (or EVENT/WINDOW), which will store

the top-level window face which is the entry point for host os events.
So EVENT/FACE is then free to be set to the face that is the
target of the event, as it was originally intended to be.


OR, instead of the above, and achieving backwards compatibility, 
you could:

Add EVENT/TARGET-FACE, which stores the target face.
(So EVENT/FACE stays as it is.)

any comments ?

I think it should be quite useful in wake-event to know which face 
an event will go to. Filtering possibilities increase.

even/face would be enough, no?