[haXe] Re: bug in neko.net.ServerLoop?

[Nicolas Cannasse]

> hey there nicolas. i am trying to track down this bug and its very difficult to
> pinpoint. basically, it seems like at some point, the length of the data i read
> is different than what the buffer has been set to and it crashes the app.

This is normal, but it doesn't crash, it justs throw an exception
because you're returning more readed-bytes than available-bytes.

> i did a test sending a lot of data to the server (i don't think i surpassed max
> buffer size) and the server crashed pretty bad and corrupted the machines memory
> to a point where it needed to be rebooted. not wure where in the app this
> occured but there should be some safeguards in the serverloop/sockets/copyBytes
> code for these things i would assume.

There are safegards, such crashes doesn't come from Neko which can't
corrupt the machine memory in any way... I guess you simply sent so much
data that the memory went too much high and your computer started
swapping. Server should prevent such a thing that by having a
max_buffer_size but I'm not sure your client test did.

Nicolas