attrs.filter differences between FR 2.2.0 and FR 3

[Palmer J.D.F.]

> Yes, this appears to of been missed off from the virtual attributes hack in the
> xlat parser.
> 
> Added it back in.
> 

Thanks Arran.
Had trouble building the git clone, so copied the code into the stable release's source, complied it and it works.

[root at server freeradius-server]# make
UNIT-TEST rfc.txt
./build/bin/radattr: symbol lookup error: ./build/bin/radattr: undefined symbol: _fr_cursor_init
make: *** [build/tests/unit/rfc.txt] Error 127


> If the request has a proxy response associated with it and no reply code set,
> this will return the proxy response.
> 
> 	case PW_RESPONSE_PACKET_TYPE:
> 	{
> 		int code = 0;
> 
> 		if (request->proxy_reply && (!request->reply || !request-
> >reply->code)) {
> 			code = request->proxy_reply->code;
> 		} else if (request->reply) {
> 			code = request->reply->code;
> 		}
> 		return talloc_strdup(ctx, fr_packet_codes[code]);
> 	}
> 
> 
> If it has a reply code set, it'll return that. Should work as expected in most
> situations.
> 
> IIRC key had to be set for this behaviour in 2.x.x. I think <response
> type>.<realm> was just a neat hack I suggested in the JRS docs for the
> Sussex case study, and other people started using it.
> 

Ah, yes that rings a bell; confirmed by blowing the dust off the case study you did all those moons ago.

Cheers,
Jezz.