[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: brain dump

Hi Aris,

It seems that this is related to my issue (http://dev.libssh.org/ticket/29).
I will try to provide you a small test case next year.

2009/12/24 Aris Adamantiadis <aris@xxxxxxxxxxxx>

> Hi Andreas,
> What gmoonen describes is not possible because each channel has a window
> size, which is updated on user read (channel_read) and not just
> channel_poll. So, this behaviour is localy possible (on small sizes) but
> does not expand to megabytes of data.
> Regards,
> Aris
> Andreas Schneider a écrit :
>  [17:18:53] <gmoonen> Owe and one more small issue i found in
>> ssh_handle_packets() if you have a slow computer (or select for detection of
>> data) and a fast network you can get a almost inf loop when doing a command
>> which receives alot of data (like cat /dev/zero)
>> [17:19:12] <gmoonen> bacause everytime a packet has been processed there
>> is already data available on the network socket.
>> [17:21:12] <gmoonen> Event I see at my local code: poll says bytes
>> available, packet_read reads it, packet_translate translates it, and
>> session_parse does it's work, and the poll function already sees more data
>> again
>> [17:21:31] <gmoonen> Command i'm using is "while true; do echo 'xxxxxxx';
>> done;"
>> [17:21:56] <gmoonen> I tell because the enter (\n) seems to split op the
>> different packets which are available [17:22:24] <gmoonen> Every poll there
>> is 8 or 16 bytes available (for data, those xxxxxxx)
>> [17:22:41] <gmoonen> (There is more available, but that is the result data
>> size)


Re: brain dumpAndreas Schneider <mail@xxxxxxxxxxxx>
brain dumpAndreas Schneider <mail@xxxxxxxxxxxx>
Re: brain dumpAris Adamantiadis <aris@xxxxxxxxxxxx>
Archive administrator: postmaster@lists.cynapses.org