Re: [wxhaskell-users] Some WX questions - a couple of missing bindings and a segfault

[Jules B. <ju...@je...>]

Eric Kow wrote:
>>  So somehow, sometimes, we're getting an invalid pointer. Whether it's a
>>  pointer to a buffer which is too short, or a buffer which has been
>>  moved, or....
> 
> Another observation is that this fails much more easily with bigger buffers.
> 
> A sort of silly question: do you think you could cook up an equivalent
> demonstrator in C++ just using wxWidgets?

Yes.

But it doesn't segfault :(

I have reduced the haskell code to the following simplified version:

import qualified Graphics.UI.WX as WX
import Graphics.UI.WXCore.Image
import Graphics.UI.WXCore.Draw
import Graphics.UI.WXCore.WxcClassesAL
import Graphics.UI.WXCore.WxcClassesMZ

import Foreign
import Data.Word

import Control.Monad

main = do
   forM_ [0..500]
     (\i -> putStrLn ("Iteration : " ++ show i) >> createAndGetPixels)

createAndGetPixels = do
     image <- imageCreateSized (WX.Size 256 256)
     pixels <- imageGetData image
     putStrLn $ "pixels : " ++ show pixels
--    putStr $ "Reading byte.."
--     forM_ [0..256*256*3] $ \b -> do
--       byte  <- peekByteOff (castPtr pixels) b :: IO Word8
--       putStr (show b ++ "..")
     bytes <- peekArray (256*256*3) (castPtr pixels)
     putStrLn $ "bytes : " ++ show (length (bytes :: [Word8]))


(The commented out code was me experimenting to see if there was a 
pattern in "which" byte causes the segfault. There isn't, that I can see)

And here is my attempt to replicate this exact set of calls from C++:

#include <iostream>

#include "wx/wx.h"
#include "wx/image.h"

void createAndGetPixels();

int main() {
   for (int i=0; i<500; i++) {
     std::cout << "Iteration " << i << std::endl;
     createAndGetPixels();
   }
}

void createAndGetPixels() {
   wxImage *im = new wxImage(256,256);
   unsigned char* pixels = im->GetData();
   std::cout << "pixels : " << ((void*)pixels) << std::endl;
   void *bytes = malloc(256*256*3);
   memcpy(bytes,pixels,256*256*3);
   //free(bytes);
   //delete im;
}


And it doesn't seem to segfault, ever.

If you leave the "free" and "delete" in then the nice deterministic C 
malloc gives you back the same block of memory each time! that's why I 
commented them out.

Jules