obexapp get failure

[Maksim Yevmenkin]

On Fri, Oct 29, 2010 at 4:01 AM, Iain Hibbert <plunky at galant.ukfsn.org> wrote:
> On Thu, 28 Oct 2010, Maksim Yevmenkin wrote:
>
>> > One thing that I can't check, what would happen if the "obexapp client GET
>> > -> remote" session get response did contain a Connection ID already,
>> > would we send it twice?
>>
>> i'm not sure... i'm tempted to say no because header are parsed, i.e.
>> we effectively move headers to another list (that is why we need
>> reparse call). however, this is something that openobex library does
>> internally, so, we need to test it.
>
> I thought about this and maybe it is a little clearer .. the object that
> is handed to the stream read/write functions is in fact the base GET/PUT
> command that openobex thinks we should be sending.. so it should be up to
> us to add any Connection ID headers since openobex doesn't know about them

i agree, openobex does not seem to know (or particularly care) about
"connection id" header. clearly, its up to the application to set it.

>>  would be really nice to try it again apple mac os x. those guys usually
>> do things according to standards :)
>
> obexapp works against Mac OS X 10.6 (see 2.5Mb log of get and put with
> connections from either side at www.netbsd.org/~plunky/obexapp-macos.txt)

thank you. very interesting. it appears that for PUT "connection id"
header only appears on the very first request. PUT-continue does not
seem to have it. it appears that mac os x consistent with obexapp here
:)

GET, however, is different, we can see "connection id" header on both
initial GET request and GET-continue requests. again, mac os x seems
to be consistent with patched obexapp. also, GET-responses on both mac
os x and patched obexapp do not contain "connection id" header.

so, it looks like GET-continue request is the only odd case here. i
wonder if this is basically wm6 obex server oddity/bug/feature :)

> I also tried my wm6 phone against it and it works fine too but I don't
> have a packet sniffer on the mac

well, yes, mac os x does put "connection id' header onto GET-continue
requests.

basically, i think, the patch should be safe.

thanks for the help Iain!

thanks,
max