[Live-devel] JPEG Video RTP Source ignores quantization table precision

[Joshua Hitchen]

Ahh yes, sorry, I should have paid more attention when writing that. Honestly it was a suggestion as a temporary patch till I get the time to write a 16 bit table support patch, which is why I rushed it.
The numtables variable was for future support (as was the way of calculating the increase to curLen).
The off-by-one error in the for loop was just negligence, but is an easy fix, the same goes for the incorrect check on the if statement.
The reason that I use the for loop rather than discarding if the entire precision field is not 0 is that there is the possibility of the excess bits not being zero - this is why the RFC states that excess bits *must* be ignored.

Other than that I fully understand your reasoning for keeping it as-is, although some documentation of the lack of support could be a good idea.


Regards

Josh Hitchen

-----Original Message-----
From: live-devel <live-devel-bounces at ns.live555.com> On Behalf Of Ross Finlayson
Sent: 09 March 2018 16:08
To: LIVE555 Streaming Media - development & use <live-devel at ns.live555.com>
Subject: Re: [Live-devel] JPEG Video RTP Source ignores quantization table precision

> unsigned curLen = 0;
> unsigned numtables = 0;
> for (int i = 8; i >= 0; i--) //each bit in the Precision field 
> indicates value of a table read from right to left; 0 for 8 bit, 1 for 16 bit (see RFC 2435 section 3.1.8) {
> 	if(curLen >= Length) break; //Ignore excess bits after all tables are accounted for.
> 	numtables++;
> 	curLen += 64 * ((Precision & (1 << i - 1) + 1); //Set up for when 16 bit tables become supported. Equivalent to 64 * ((Precision & 2^(i-1)) + 1).
> 	if((Precision & (1 << i - 1) == 0) return False; //Currently unsupported - 16 bit precision table.
> }

Given that there seems to be so at least three things wrong with this:
	- i ranges from 8 through 0 (i.e., takes 9 different values)
	- The variable “numTables” is assigned but not used
	- You’re checking bits in the “Precision” field to be 0 to mean a 16-bit precision table; you should instead be testing for 1-bits I’m not going to be adding this to the code, thank you :-)

In any case, if we were ever to support 16-bit precisions in a quantization table, then we’d need to completely change the “createJPEGHeader()” function - to take the “Precision” field as parameter, and use it in the implementation.  If and when someone proposes a patch to do that (not I, because I don’t think people should really be streaming JPEG anyway :-), then I’ll consider adding it to the code.  In the meantime, all we could really do is just check whether any 16-bit tables are present, and, if so, reject the incoming RTP packet.  That’s what your code aims to do, but a much simpler way to do that would simply be to do
	if (Precision != 0) return False;
instead of your code.  But I’m not going to do that either, because it will cause a client who tries to receive/handle 16-bit tables to just fail silently.  Instead, the current code will cause a client who receives 16-bit tables to render an incorrect (i.e., bad-looking) JPEG image, which would better help them figure out what’s wrong.


Ross Finlayson
Live Networks, Inc.
http://www.live555.com/


_______________________________________________
live-devel mailing list
live-devel at lists.live555.com
http://lists.live555.com/mailman/listinfo/live-devel