AltME groups: search

about #4172 fix in  2.7.2... YIPPE! 


can you imagine that this has been there for years! just show how 
some things can slip through the cracks.  it prevented a few faces 
from being pre-set within VID dialect.

Can someone show me an example of how center-face could now be used 
to center off a parent face?

henrik, maybe it's just a coincidence...

I believe the PE file format has sections that are padded, so it 
may be that the internal padding is different, as are the code segments, 
but the resulting EXE is the same size. I suppose you could do a 
binary diff to see.

Also, in the editor window, hitting the  'home' key produces a null 
character, and 'end' does nothing.

my rotation zoomer finally works under OSX :-) A dream come true. 
but the colors are screwed!

or by timers is carl just referring to the VID based rate/time events 
rather than a Core accessible timer?

that blog was probably provoked by my earlier comment in the blog 
... at least Carl confirmed, that 2.7 will see only mostly those 
changes, which could be used as a codebase for 3.0 too ...

I'm not sure what you mean by slipping. I wouldn't expect 3.0 to 
be usable for production work within a year anyway (I think I mentioned 
that before)

so holding off 3.0 for a few weeks to get 2.7 out, gets my approval, 
particularly if some of my bugs are fixed :-)

I did not expect it to come in a year. If so, then of course 2.7 
still makes a lots of sense ...

I'm of course expecting an R3 alpha to come out sooner than a year, 
but I wouldn't expect or rely on having a production level R3 within 
a year. If it comes sooner than that, then fine. :-)

Desktop Help window still says Copyright 2001 - 2005. Maybe it should 
be a global string in system?

I think doing all these new builds is a good thing in at least one 
regard; if it gets RT into the habit of doing quick builds and releases, 
it will be much better when early R3 builds come out. That is, if 
it becomes habit, and the tools are there to make it easy to deploy 
new builds, we'll make much better progress testing R3.

I'm just afraid this behaviour will fade out just like every promising 
thing in the past (RT Q&A thursday and so on).

I've posted this bug in the comments to 2.7.4 OSX build, but I'll 
repeat it here, in case it's being seen here first. It's a colour 
problem, when transforming in the DRAW dialect. Code example:

i: make image! [100x100 255.0.0]
draw i [pen green line 10x10 90x90]

view layout [box effect [draw [image i transform 0 -30x-20 2 2 0x0 
image i]]]

There seem to be a problem with read/binary. Maybe it's the same 
bug!?
read/binary http://www.rebol.com
returns empty.

Maybe you should report it as a comment in his blog! I think, he 
sees it then.

it was actually reported as a blog comment about 2 weeks ago

Btw... Calr spoke about a different compiler used for Windows version, 
what C compiler is that? , just currious...

Does line-join in the DRAW dialect works with Windows version? There 
seem to be a problem with OSX version:


view layout [box 200x100 effect [draw [pen red line-width 15 line-join 
round line 10x10 90x10 90x90 10x10 polygon 110x10 190x10 190x90]]]

I think, they both should be with rounded corners.

Geomol can you give the link to canvas RPaint... I have a very interesting 
proposition for you.  but I need to look at RPaint before   :-)

hum I found a link to version 0.9.8 but on execution I get an error: 

** Script Error: Cannot use path on none! value
** Where: repend
** Near: system/script/header/version

that's with v1.3.2 of REBOL.

using 2.7 I get a not enough memory error and can't even start it.

I remember that Geomol wrote a version with a fill-pen that was written 
in rebcode. it's very fast, practically instantanous.

wow.  I hope rebcode gets inserted within 2.7 release.  THAT would 
be a differentiating feature.

:-)

Maxim and Henrik, I took screen-dragging out again, because it didn't 
work 100%. I think, the code is still in there commented out.

Maxim, if you have a proposition, please feel free to write me privately.

Geomol: The behaviour of LINE you have spotted is not a bug. Try 
this code to see the difference:

Hmm, when comparing 1.3.2 and 2.7.4: The LINE-CAP round has no effect 
in 2.7.4?!? Looks like a bug.

Geomol, to explain it a bit:
LINE - draws single lines (eve if you specify multiple coords)

SPLINE - draws 'connected' lines but the 'head' and 'tail' is splitted 
(until you specify CLOSED keyword)
POLYGON - always draws 'connected' lines in form of closed shape

that' probably a endianess issue.

SPLINE actually seems a little faster than LINE. Interesting! :-)

It can be faster because the SPLINE is drawing line in a loop while 
LINE is called separately for each line.

Cyphre, my goal is to get Canvan RPaint for version 1.0 and work 
the same on Win, OSX and Linux. Then REBOL/View is in a condition, 
I'll call good! :-)

AFAIK Linux has the DRAW text support added(using fretype2 library). 
The only problem is how to get path to a font on Linux.(as there 
seems to be no 100% 'standard' way how to get it on all distros we 
need to use yet another dependency :-/ )

cyphre, has the window-to-front fix for OSX from the blog been noticed? 
someone posted a bit of code to eliminate that problem.

cyphre: oh, Carl is talking about a focusing problem, but it would 
be the same thing: http://www.rebol.net/cgi-bin/blog.r?view=0157

btw font handling paths in linux is a problem for all apps... not 
just REBOL.

why not just have that extra parameter setup as a global word ?  
then it can be setup within user.r if need be.

(looks like a typo to me)

I seem to remember, that holding <ctrl>- or <shift>-key under Windows 
while pressing a mouse-button would be detected. Can anyone verify 
that? I can't get it to work under OSX.

I'll post a comment about it.

well, in a way they do. there is a ctrl, alt and cmd key. on PC keyboards, 
the ctrl key handles things that normally is assigned to the cmd 
key and the ctrl key. this makes it a bit confusing to copy (ctrl-c) 
things from a terminal on a PC keyboard, whereas the mac keyboard 
uses cmd-C to copy. in rebol this means that the ctrl functionality 
in windows should be mapped to cmd on the mac, but if this was to 
be done right, we'd need a /cmd refinement as well.

Oh, the cmd-key (apple-key) will actually set event/control true, 
but only for key events. As it is now, I don't think, you can check 
for control or shift when getting a mouse event.

that was a long week ;)

pekr, yes there is probably a little dilemma there.

I'd really like if rebcode where included in 2.7 but had to be allowed 
within secure... so it would not be a security issue unless you want 
to use anything that you trust is ok.

all it would need is support for struct and arrays.

I feel like I  could really optimise a lot fo my code for it.

when using rebcode?  or 2.7?  so far I've had very few real problems.... 
even within elixir which has a lot of processing going on.

Well, that depends on the workload for Carl. If they have to test 
R3 stuff in R2.7, I'm fine with that. Besides I suspect that R2 won't 
be going away until at least a year from now. We simply don't know 
whether R3 will be stable enough in its first incarnation for production 
use. R2 is stable enough as long as you stay with smaller and simpler 
things that aren't graphics heavy. I tend to bend that rule a bit. 
:-)

but a part from async http closing of ports, I have not had a single 
crash of rebol in many months... a year? maybe.

it depends... retask is a very graphics heavy app and it can run 
for days without being ended.

but most apps I use are punctual in use, but that will change with 
elixir in a few weeks

altme can run a long while too... without crashing.

About crashes, I submitted 4267, 4238, 4167, 4137, all crashes, and 
I found them on my own. Besides I've bumped into recycle crashes, 
port crashes/hangs and also other crashes, that I can't put my finger 
on, when using Rugby with secure network transfers. I get a crash 
about 1-2 times a day.

yes... there is something weird with the port management.  as I said 
a part from that I've had no bug a part from a few AGG problems.

I did put a word with Carl about the tcp port instability of R2 at 
one of the informal late nite meetings some of us had. Gabriele agreed 
to some extent that there was some port strangeness so if that gets 
any real overview by Carl probably remains to be seen, but at some 
level he knows we are having some oddities with it.

4267, 4238, 4167 all seem reasonably easy to work around Henrik; 
not that I'm saying they aren't important, or shouldn't be fixed, 
but we don't know RT's internal development setup, so we can't really 
say what they should do. We all have different priorities. 


I would like to see a large set of public tests, to help RT nail 
bugs and catch them when they sneak in.

I still have a Rugby app that runs away once in a while, too. I don't 
know if it's Rugby, REBOL, or something on that specific machine, 
because it never did it before this new machine was set up.

please guys keep in mind we can't fix all bugs at once. and we can't 
usually fix two "big" bugs at once (big means it takes a lot of time 
for Carl to figure it out). so we fix one big bug and many small 
ones (bugs that can be fixed in a couple minutes always have precedence 
because there's no point in making them stay for years on rambo).

also... "sometimes rebol crashes" is something that's not going to 
get fixed. there needs to be a way to reproduce the crash - otherwise 
carl would need to spend many days on it.

Then again, it isn't always easy to lift the bug from  [long running 
 / large volumes of data / 1000s of lines of script] applications, 
and set it into a small example.

I once tried to produce a small script that relicated the block-manipulating 
logic of an app I have. I was hoping that running it over night would 
create a "should not happen" bug. It never did.

By way of compensation. I did find this one-liner that will crash 
(without warning) every version of REBOL I've tried it on:

     a: copy [] b: copy [] insert/only a b insert/only b a same? a b
(I think I reported that, but it was before RAMBO)

You will fix the Linux side right?  "Sometimes rebol crashes".  Sorry. 
 :)  I'm still 
attempting to find a short piece of code that trips the segfault.

i'm not saying the crash bugs should not be fixed. i'm saying that 
in the current situation there are not enough resources to fix them 
unless there's a simple way to reproduce them. hopefully, after r3 
is out, there will be more time to give to r2 and we'll be able to 
fix those misterious bugs too.

Is it easy to switch off gc inside a wait?

Thanks, Gabriele.....My collection of REBOLs are a point release 
or two behind

I started using 2.7.6.

It's great to be able to "call" external apps without the dos window 
popping up. I have alot of use for this. 

I noticed, however, that when calling wmplayer in windows the file 
will play in the background which I also find appealing but there 
are times I might want wmplayer to open. Is there a way to force 
this with call?

There does seem to be a problem with 'call that wasn't there before.

regarding your example in R2-Beta - Bugs (Problems that need solving.)

This seems to be a windows behaviour. It works with the complete 
path;
 call "%windir%\explorer.exe C:\"

Not sure, but maybe they're trying to get a console handle, and can't.

i put this on rambo a while ago (same reason, imap servers)

Not a problem.  Just here the country is out of wack.  Justices are 
now making laws.

My guess: it calls via the shell (cmd under e.g. WinXP) instead of 
a directly Windows call.

There's a page on differences between 2.6 and 2.5

Is anyone else having this occur:

When ever I start a script now I get an empty "public" folder created 
in that directory.

Little bit more playing around, to see if +s was taking ... doesn't 
seem so.
rebol []
probe system/script
change-dir %..
list-dir
ask "waiting"

c:\tools\rebol\appdata\rebview +s testing.r

...
    words: none
]
waiting

** Script Error: list-dir expected dir argument of type: file url 
unset
** Near: list-dir
ask "waiting"
>>

The ask is evaluated before the list-dir error     And if  list-dir 
%.  is used, it lists the dir, without a security box.

There seems to be a difference in CALL/SHELL (at least under Windows). 
Earlier versions open a shell window, the latest doesn't. I can see 
it, when I run my OpenGL stuff.

it's been going down every day. i have restarted it twice, but there 
seems to be something wrong with it, so i'll be waiting for carl 
to be back and have a look at it (since i have no idea how altserve 
works :)

not a disk problem, this is on www.rebol.net.

Rebolers, again I reach out to you for some info and advice. I'm 
creating a DB in Access 2003 for a company (This is what they use 
so I'm stuck here) and I'd rather create a set of tools (for some 
work but not all, the majority of the tasks will be strictly access) 
that has Rebol talking to it rather than having to doing it in VB. 
I was under the impression that 2.7 had odbc built in, but alas it 
doesn't seem to. Am I stuck with having to get  /Command?

At the moment I have no access to that old pc and script. I takes 
a few days, but then you can get it.

At least someone was even using it on Linux, as Command does not 
provide ODBC access there.

Thanks in advance then. I'm really trying hard to not have to learn 
VB : ) Which may not be a good thing in the long run. Thanks.

You should also consider having them use SQL Server 2008 Express 
instead of Access for the database portion, though you still have 
to be careful about datatype translation if you are still using Access 
as a front-end, even 2007.

Thanks Brian. I forgot about Express. If I can convince them to let 
me do a complete Rebol app I would do that. : )

Thanks and this project is pretty basic. It's more along the lines 
of a task management system. Right now they do it with some fairly 
bulky spreadsheets. Thanks for the tip.

Would you send me a copy too?

Sqlab, OK, I gave your odbc-lib a shot. So far so good. Thanks. I'll 
be doing some more testing over the weekend. If you know of any quirks, 
please let me know. I'm pretty much an "insert" and "Select" SQL 
guy so it should be fine. Thanks again.

sqlab - can I get a copy of your ODBC-lib as well?  The email address 
inmy profie is valid.

You know I thought I saw that but this morning I went to the main 
site and that's pretty much the same as it always has been.
Thanks. I'll take a look.

Is it possible to hide a button that is the button one just pressed? 
my_button: btn "test" [hide my_button]

You have to be a ltitle tricky to hide it James, because it wants 
to redraw when you release the mouse button. In the past, I've set 
up a central handler and states to control what is shown, then your 
action just sets the state you want and the button will be hidden 
on the next pass (timer event).

Thanks Gregg, that is tricky. Well, it wasn't a super important item 
so I ended up using the enable/disable. Thanks anyway.

The show option, and hiding the shell window by default, was a change 
in new releases.

Q: "Any idea why?" call/show

A: Most of the time call is used to run background processes, and 
REBOL <= 2.7.6 did not give you the option, always flashing that 
shell window. Starting with 2.7.7 the called app (and the shell window) 
is not shown by default.

No, and work on it has barely started. We will have a public development 
release of the new R3 GUI first.

Hi,
For homework, I was asked about the random number generator in 
my favorite language ... did some experiments but answered 
closed source
 to what "algorithm does it use?" for this part, I got a zero.

So, could someone with access to rebol source or internal doc 

get me information on the algorithms for 'random and 'random/secure

they are probably some flavor of linear congruent generators 
(please don't say RANDU) but multiplictive?  or mixed?  
maybe a PMMLCG? shuffling? QCG? MRG? What constants are used? 
what is the period? how many streams? 


The easiest on you would be just send me the source and let me figure 
it out 

but is that is verboten as much detail as you may send to me (privately)
would be appreciated 

note: the information will only be seen in a distilled form by my 
professor
and not kept by me.

thanks Tom

Hi Tom, I think, that the best what can be done is to put this to 
CureCode (R3 uses the same algorithm, I guess) as a documentation 
bug

Random: I have a simulation, where I use www.random.org data to make 
sure I am truly random

some things are stunningly non-obvious when it comes time to optimize 
REBOL


>> a: 1x1 b: 10x10 s: now/precise loop 1000000 [second b - a] difference 
now/precise s
== 0:00:00.5


>> a: 1x1 b: 10x10 s: now/precise loop 1000000 [b/y - a/y] difference 
now/precise s
== 0:00:00.969

all path access is slow...but I woudn't have thought that using a 
series function and multiplying both values in the pair would be 
twice as fast!

Went through the Parse Proposals list and tweaked the Priorities 
section to double as a status list. Unfortunately, some of the rejected 
proposals must have at some point been removed from the proposals 
page. We wanted to document them and why they were rejected, so they 
don't get proposed again. I may have to go the history and find the 
rejected proposals that were deleted and restore them to the page, 
so they can be rejected explicitly with explanations.