From: Jean T. <jt...@bo...> - 2002-09-24 02:05:39
|
On Sat, Sep 21, 2002 at 11:51:43AM +0100, Gareth Bushell wrote: > On Friday 06 September 2002 03:10 am, you wrote: > > Gareth Bushell wrote : > > > The obex continue response can contain data but the current library > > > flags this as an error. This interfears with a protocol layered ontop > > > such as syncml. > > > > I'm trying to put up a new release of OpenObex, and I was > > looking at your patch. And I've got troubles with it. > > > > Obex is supposed to be straighforward clone of HTTP, so I > > really don't understand where this stuff would fit. You can receive > > data in continue properly when you are receiving data, i.e. when you > > have finished sending your request. I've tested the GET support many > > time, and it works properly. > > So, you seem to say that the server sends back data before we > > have finished submitting a request. That seems totally wrong. > > Furthermore, I don't understand what we are supposed to do > > with this data. > > > > So, would you mind sending a log/dump of the Obex > > transaction. If you are using Linux-IrDA, just use irdadump. > > > > > > Hi > Sorry about the delay, I have used irdadump to log the upload of the phone > book to my T68 mobile phone. This is when data is contained in the continue > message. The irdadump utility SEGVs on the first part of the upload and I > ran it again after it have crashed and these are the two log files attached. > There is a gdb run of irdadump when it SEGVs but It doesn't look to helpful > to me. ( gdb-irdadump.txt). Grr... irdadump should not do that (obviously). And when you restart it, it doesn't pick up that it's Obex, so doesn't do the proper decoding, which make it pretty useless. You are using an old version of irdadump. But I'm wondering what is making irdadump choke. I think it's related to the connection response, and I remember messing up with hat in irdadump at some point. You would need to compile it with debugging (-g) to allow gdb to do its job. Probably I should add an option to disable decoding, that would allow to get more meaningful bug reports... > Also included it output from my program t68tool > that prints a dump of the continue data. (t68tool.output) You can get T68tool > from www.garthy.com. Actually, that was pretty good. But the mystery thicken. The header it's returning is supposed to be used for authentication during the connect, and not at all at that point, so I suspect they are violating the protocol (typical for Ericsson). And they change it at every frame. I don't see how we can handle that sanely. Yuck ! > Anything else I can do to hepl you and get this patch in openobex (or > something better as i know I don't do the right thing in the patch but it's a > temp fix) I'd be happy to do. I need to think about it. I think your fix will be the right one, as we can't do much else than dropping the incomming data. > I'm using the debian unstable version of irDADump. > > irda-utils (0.9.14-17) unstable; > > Kernel version > > # cat /proc/version > Linux version 2.4.18-k7 (herbert@gondolin) (gcc version 2.95.4 20011002 > (Debian prerelease)) #1 Sun Apr 14 13:19:11 EST 2002 > > > Cheers > > Garth Good luck... Jean |