justnajm / physaxe Goto Github PK

The attached source produces a swf which runs for 15 seconds and then
crashes with this error message:

Error: Error #1502: A script has executed for longer than the default
timeout period of 15 seconds.
        at phx::World/onCollide()
        at phx.col::SortedList/collide()[./phx/col/SortedList.hx:85]
        at phx::World/step()[./phx/Const.hx:184]
        at Test$/loop()[./Test.hx:29]

Found on Linux using haXe 2.03 and Flash Player 10 with Physaxe svn source
from what seems to be April 18th, 2009 (I removed the svn tags and don't
remember, so I just ls -l 'd it).

What steps will reproduce the problem?

"haxe -swf test.swf
-swf-version 9
-swf-header 400:300:30
-main Test
-lib physaxe
--next
-main Test
-lib physaxe
-js Test.js"
to compile Test.hx as suggested in Tutorial at
http://code.google.com/p/physaxe/wiki/Tutorial

What is the expected output? What do you see instead?

sohould be compile without failures in library, instead the follwoing error
occured:
"/usr/local/lib/haxe/lib/physaxe/1,2/phx/Body.hx:148: characters 8-15 :
Unexpected dynamic"

What version of the product are you using? On what operating system?

haxe (Haxe Compiler 1.19), packaged from Debian lenny (1:1.19-3)


Please provide any additional information below.

physaxe was installed with help of "haxelib install physaxe".

What steps will reproduce the problem?
1. Create a body.
2. Create a Polygon shape and specify the vertices in clock-wise order.
3. Add the shape to the body, add the body to the world.

What is the expected output? What do you see instead?
The created polygon should react to collisions, but instead completely
ignores them (this includes manually checking if it contains a point that's
obviously inside the polygon).

What version of the product are you using? On what operating system?
Physaxe 1.2, Windows XP.

Please provide any additional information below.

[deleted issue]

Add clone methods to the Material and the Properties classes.

As these classes are mutable, and instances of these classes are often shared, 
clone methods are desired.

Patch attached. Hit me up if you have any comments (on the implementation, for 
instance).

What steps will reproduce the problem?

1. move with the arrow keys, space sets velocity to zero
2. let the circle fall a sleep
3. press right arrow, I set body.island.sleeping=true so I could adjust the
velocity again, but it falls thru the floor in the process

What is the expected output? What do you see instead?

The idea of idle parts of the world falling asleep is a great optimization,
but when I try to programaticly it up, the model seems to get confused.

What version of the product are you using? On what operating system?

I'm using haxe 2.01, compiling to flash 9 on Linux.

Please provide any additional information below.

Here is how I tried to wake up the island:

 if ( bob.island.sleeping ) {
    bob.island.sleeping = false;
    bob.v.y--;
 }

I tried to pop bob up a bit, but it didn't help.

Sometimes to make it fall asleep, I used SHIFT + arrows (moves
flash.Lib.current.x and flash.Lib.current.y)

I expect I may not be using the library correctly.

Altered syncShape so that rather than searching from the start of the list to 
find its place it now searches from its current location. For high-movement 
problems a large proportion of the shapes can potentially be repositioned by 
this function and hence it is important that it is algorithmically efficient. I 
can think of few general cases where this patch would degrade performance.

The effect of this patch using demo 5 (falling pentagons) was a decrease in 
total execution time per iteration (the "all" timer) on my machine to 65%-75% 
of pre-patch values (ie - 25-35% faster). This is obviously a significant 
performance gain.

What steps will reproduce the problem?
1. create a shape with restitution 1 and add it to a body
2. test it to collide with other body

What is the expected output? What do you see instead?
the expected output : the body will bounce elastically
i see instead : the body doesn't bounce elastically

What version of the product are you using? On what operating system?
physaxe 1.2
haxe 2.05

Please provide any additional information below.
if i change the step that i pass to world.step() function, it behaves 
differently.
the bigger the step, the more bouncy the body

As stated by Hugh on
http://code.google.com/p/hxcpp/issues/detail?id=90

The use of Math shortcut will create a huge performance hit on c++ target.
The only use of this technique in physaxe is 
http://code.google.com/p/physaxe/source/browse/trunk/phx/Island.hx#81

Linear friction is inverted.

The docs say "  The amount the linear speed of the object is reduced by
time", implying a higher number == higher friction.

However, Island.hx show the opposite:
v.x = v.x * p.lfdt + (g.x + b.f.x * b.invMass) * dt;

This is confirmed with a simulation, and the fix is easy.

I know documentation is always needed, but at the minimum, the
accepted range of values should be given.  Actual friction
coefficients can be > 1, but in this case, a friction coefficient > 1
results in additive velocity!!!

I volunteer to add some documentation as I use this code, if I'm
granted permissions.

Useful for detecting if an object is inside a specific region.  The
simplest way I see to do this is by adding these few lines to prevent
non-physical bodies from running the impulse/force calculation code.

Body.hx
< public var physical : Bool;

public function new()
< physical = true;

Arbiter.hx
public function applyImpulse()
< if (!b1.physical || !b2.physical)
<           return;
public function preStep( dt : Float )
< if (!b1.physical || !b2.physical)
<           return;

This seemed to be working fine with 2.04, but 2.05 has problems.

You get the following error:

uncaught exception: mysql_affected_rows(): supplied argument is not a valid
MySQL-Link resource (errno: 2) in
C:\DEVELOPMENT\lightcreative_arenaint\www\lib\php\db\_Mysql\MysqlResultSet.class
.php
at line #14mysql_affected_rows(): supplied argument is not a valid
MySQL-Link resource

in file:
C:\DEVELOPMENT\lightcreative_arenaint\www\lib\php\db\_Mysql\MysqlResultSet.class
.php
line 14
#0
C:\DEVELOPMENT\lightcreative_arenaint\www\lib\php\db\_Mysql\MysqlResultSet.class
.php(14):
_hx_error_handler(2, 'mysql_affected_...', 'C:\DEVELOPMENT\...', 14, Array)
#1 C:\DEVELOPMENT\lightcreative_arenaint\www\lib\Test.class.php(6):
php_db__Mysql_MysqlResultSet->getLength(true)
#2 C:\DEVELOPMENT\lightcreative_arenaint\www\lib\Main.class.php(8):
Test->__construct()
#3 C:\DEVELOPMENT\lightcreative_arenaint\www\index.php(11): Main::main()
#4 {main}

when trying to do an UPDATE operation.

What steps will reproduce the problem?

var cnx = Mysql.connect([details]);
var affected = cnx.request("UPDATE `test` SET `var`='y' WHERE id=1").length);
trace(affected);

What is the expected output? What do you see instead?

An integer of how many rows are affected.

What version of the product are you using? On what operating system?

2.05 / Windows (Vista 64) / PHP 5.2.6

Please provide any additional information below.

I found I could fix it by changing a line in the library MysqlResultSet
class in php/db/Mysql.hx file. You just have to get rid of the result set
itself in the following line:

return untyped __call__("mysql_affected_rows", __r);

and change to:

return untyped __call__("mysql_affected_rows");

I've had this problem before when doing straight PHP, that if you give an
link resource to this function it complains. Leaving it out means it just
uses the last one used. I guess my fix would cause hard to find errors for
people who are connecting to more than one database so another fix should
probably be found.

What steps will reproduce the problem?
1. Create an object that will be moving very quickly across a different object.
2. Run.

What is the expected output? What do you see instead?
The fast-moving object will skip over the boundaries of the other object,
rather than bouncing off.

What version of the product are you using? On what operating system?
1.1, haXe 2.0.1 on Linux in Flash.

Please provide any additional information below.
This causes problems when moving objects happen to fall through boundaries
of others, I believe. It may only happen with circles? See the attached demo.