[Live-devel] Motion JPEG header reconstruction bug

[Massimo Corà]

Hi,

I collected some other debug-infos:
changing the qFactor value determines differents outputXX.jpg views, like in

http://www.dei.unipd.it/~coramass/livemedia/

where XX is the referring qFactor. As you can see only the image version
with qFactor = 130 does not suffer of the 4-btyes problem. On the other
side with value 90 we have the nearer image to the original one. Huffman
tables seems broken, perhaps because of the "bug"?

Massimo Corà wrote:
> Now the problems I encountered: first of all, FileSink doesn't write
> anything to the output file with a calling like
> FileSink::createNew(*env, "jpeg_received.jpg", 60000);
> To have some data written I must set the file to stdout and then from
> cmd line call: jpeg_receiver.exe >> output_file.jpg


uhm I think I missed here. Problem was that I didn't wait enough time
for the receiver to start up, sorry.
Anyway, is there a method to have the jpeg-frame pointer to memory
without flushing it to disk and reload? I cannot find one..

thanks and regards,

Massimo Corà