[0.4.0]download disconnects on $adcget

[braindancer 0]

first of all, nice to see a peerguardian plug-in

the problem i'm having is this:

i have a download running, and every 90 seconds it disconnects for no apparent reason.

i checked the logs and it seems that it is trying to initiate another download of the same file:

<snip>

Client: [Outgoing][10.0.0.2] $ADCGET file TTH/[hash of file]

Client: [incoming][10.0.0.2] $ADCSND file tth/[hash of file]

</snip>

90 seconds later

<snip>

Client: [Outgoing][10.0.0.2] $ADCGET file TTH/[hash of file]

Client: [incoming][10.0.0.2] X/?4u+$<J=k

</snip>

*poof* disconnect.

i also noticed that the [C] tag is not showing on my progress bar,

so i assume the file is not being downloaded in chunks. if so, why?

and how to force chunkage? and how to set chunk size?

regards, braindancer

-edit-

the other client is a dcgui/valknut 0.3.8

[Zlobomir 3]

Erm, could you try and reproduce this scenario with a DC++/SDC++/ApexDC++ client on Windows at the opposite end? Just to ensure it is not Valknut's fault.

[braindancer 0]

can't atm, since i'm in a vpn and i have the only windows box

some piece of hardware in this box is unstable, and until that's

fixed, it's no use even trying to run a sourcebased distro on it.

what i was wondering is, why the second $adcget?

to download a second segment at the same time?

on another note: another user was able to download from me

without any problems, [Z] and [C] worked. he was using valknut too.

so i guess it's that second $adcget that's doing no good.

regards, braindancer

-edit- having the exact same problem when downloading from a different valknut.

i noticed that before the second $adcget the client does not issue the

$lock extendedprotocol[addmorehere]

if it'd help, i can provide bigger log snips,

so someone who Knows What He Is Doing can compare the requests.

-editedit-

we may be able to test this with another win client. a friend just offered

to run a vmware

-yet another edit-

on closer inspection...

the initial file request has a header like this:

<snip>

$mynick [local nick]

$mynick [remote nick]

$lock [local, value]

$lock [remote, value]

$supports [local, value]

$direction [local, value]

$key [local, value]

$supports [remote, value]

$direction [remote, value]

$key [remote, value]

>$adcget file tth/[hash]

<$adcsnd tth/[hash]

</snip>

download starts. 90 secs later, without any header at all,

my client issues

$adcget file tth/[hash]

to which valknut responds with apparent garbage

[incoming][10.0.0.2] JNKIQ3

[braindancer 0]

so we did some testing, and it appears that the adcget implementation of valknut is at fault.

on an initial request, apexdc sends the headers i listed above, and then the initial adcget command.

valknut understands that and responds adcsend.

then after an arbitrary amount of time (i guess after a chunk/segment is done),

apexdc issues a second adcget, this time completely without headers.

since i am fairly new to apex, i can only guess that this is a 'next chunk/segment' or 'continuation' request.

but valknut seems not to understand this 'headerless request' and responds with garbage,

causing a disconnect.

apex does not have a problem with those 'headerless requests'

i will contact the valknut people about this, will post back here.

[braindancer 0]

ok, this topic can be closed.

ejs1920, the main valknut developer established that it

is a bug in valknut.

see http://dcgui.berlios.de/forum/viewtopic.php?t=3540 for reference

he also provided a valknut patch to remove this bug.

the patch is currently being tested and may be implemented

in valknut 0.3.9.

thanks for your input.

regards, braindancer

[Zlobomir 3]

Thanks also to you and ejs1920 for the investigation. Sorry for my late reply, but forums were offline.