Memory leak remains in 1.94

Archives Forums/Blitz3D Bug Reports/Memory leak remains in 1.94

Damien Sturdy(Posted 2006) [#1]
Confirmed: Memory leak freeing images:
Not only loaded images but created images as such:

Graphics 640,480

Repeat
IMAGE=CreateImage(640,640)
FreeImage image
Delay 1000
Until KeyHit(1)
End



I hope this is an easy fix! :D


Ross C(Posted 2006) [#2]
Me too! :o)


Boiled Sweets(Posted 2006) [#3]
Mark?


Damien Sturdy(Posted 2006) [#4]
Mark died of stress :P

Have faith, He'll fix it :D


Danny(Posted 2006) [#5]
Here's another guaranteed memory leak from hell regarding textures! The leak occures at the RenderWorld command. But if you comment out the WritePixel - the leak won't occur either.

I imagine that changing a texture will force it to be copied to the video mem again, but before that happens, some buffer overflows...
If you use the 256 flag on LoadTexture (force loading texture in video memory) the leak doesn't occur!
Also it seems only to happen on LOADED textures, when you use CreateTexture() it doesn't happen either..
; NOTE: Using Flag 256 (force load in VMEM) DOES prevent the leak!

Graphics3D 800,600, 0, 2
ClsColor 0,0,20
Color 255,255,255
SetBuffer BackBuffer()


camera = CreateCamera()
PositionEntity camera, 1,3,-3

cube = CreateCube()
tex = LoadTexture("C:\Program Files\Blitz3D\samples\mak\b3dlogo.jpg")	;--> MUST LOAD a texture, ie. CreateTexture doesn't give leak!
EntityTexture cube, tex

PointEntity camera, cube
size = TextureHeight(tex)

While Not KeyHit(1)

	SetBuffer TextureBuffer(tex)
	WritePixel Rand(0,size-1),Rand(0,size-1), $ff0ff
	SetBuffer BackBuffer()

	RenderWorld		;--> Leak!

	Text 5,10, "painting random pixels on texture, watch memory leak increase in TaskManager!"
	Text 5,30, "Press Escape to end.."

	Flip 0

Wend


good luck,
D.


DJWoodgate(Posted 2006) [#6]
I have not upgraded to the latest version, but it occurs to me that CreateTexture does not by default enable the mipmap flag. Might be relevant. Try Cleartexturefilters() before loading the texture and see if that makes a difference.


Damien Sturdy(Posted 2006) [#7]
Wow, That's an impressive report! Perhaps you hit the nail on the head :)

Mark, come save us!


Damien Sturdy(Posted 2006) [#8]
Mark, can i suggest taking down the latest 1.92,1.93, and 1.94 updates, since the pretty major bugs might and probrably will become a problem to new users who don't understand them- More so to anyone who doesn't frequent the bug report forum. I think it's in the best interest?

Plus, you can pretend they never existed afterwards! :D

[edit]
What does everyone else feel about this? Am I about to get stoned about it? :)


Danny(Posted 2006) [#9]
I totally agree Cygnus. I sounds like one of those 'little bugs - with mayor consequences accross the board' and probably one of those really hard to find one's ;)
I can't imagine these type of serious bugs would still be better than whatever 1.92-1.94 should have fixed.
And yes, I defenitly agree this could create havock amongst the new users, or at least leave a bad taste...
But even for the not-so-new-users, when you've grown to rely on blitz for doing this and that, you probably first start thinking it's your own fault causing this or that memory leak or mav :)
D.


big10p(Posted 2006) [#10]
v1.92 and v1.93 have already beem taken down. Look into my eyes - not around my eyes - look into my eyes: "They never existed, OK?". :P


Damien Sturdy(Posted 2006) [#11]
One of blitz' stong points is its bulletproof stability- (mostly). A broken Free command puts a hole in that, 1.91 is the last stable release i recon- no memory leaks, and a blitz crash meant a user mistake :)

At present there are ways blitz can fail that are non-user-mistake related.


OJay(Posted 2006) [#12]
didn't you get it? mark wants us to upgrade to max! \o/

nah, no trouble whatsoever...v1.91 is just working fine...

oh and: do you remember: "Anything that can go wrong, will go wrong." and: "Never touch a running system."


so long... :)


jfk EO-11110(Posted 2006) [#13]
I understand this is a serie of bad luck and Mark must think this whole thing really sucks. I don't mind if 1.95 is going to take a while. But 1.94 should be removed. Or maybe, to be on the save side, a copy of 1.91 should be renamed to 1.94b and released. Seriously.


QuickSilva(Posted 2006) [#14]
Any official news on this situation?

Jason.


popcade(Posted 2006) [#15]
So we have to back into 1.91 and wait the fix?

I'm still waiting for new BMax version....


jfk EO-11110(Posted 2006) [#16]
>>So we have to back into 1.91 and wait the fix?<<

Since several people reported the image memory leak I'd say: yes, this may be the best option right now.

Not shure if you can downgrade with a simple install of 1.91 since some people said 1.93 has a new linker version that won't work with 1.91, probably you need to uninstall it and then make a fresh install, be it from the hopefully backed up buy or from the original CD, and then update to 1.91.


Boiled Sweets(Posted 2006) [#17]
Some news on the fix would be nice - even if it's won't be available for a while


King Dave(Posted 2006) [#18]
Just official word that its a known problem would do, not the old ignoring tactic.


big10p(Posted 2006) [#19]
lol, I'm pretty sure they're aware of the problem. There'll be a fix along at some point, I'm sure. We'll just have to be patient, as usual. :)


marksibly(Posted 2006) [#20]
It's a known problem - fix coming soon.


big10p(Posted 2006) [#21]
Told ya! :P


King Dave(Posted 2006) [#22]
:) Thats all I wanted :P You're right I'm sure it couldn't have been missed but its still nice to hear... after all it did exist in the 1.93, so it was a surprise to see it in 1.94 too if it was known about :P