Re: [JSch-users] stress testing jsch (+ commons-vfs)

[Jerome L. <jer...@gm...>]

On Jan 9, 2008 3:45 PM, Atsuhiko Yamanaka <ym...@jc...> wrote:
> Hi,
>
>    +-From: "Jerome Lacoste" <jer...@gm...> --
>    |_Date: Fri, 21 Dec 2007 22:21:45 +0100 ______________
>    |
>    |Attached is a patch that improved the reliability of jsch. Details:
>    |* compile jsch with debug mode by default
>    |* a known issue in jsch 0.1.36
>    |* don't consider a read() that returns 0 to be a closed stream
>
> I have a question.
> I have understood your intenstion, but 'i<0' may cause the infinite busy

You mean i == 0, right ?

> loop and it must not be acceptable.
> Do you mean that you had gotten the problem with 'i<=0'?

I've fixed several bugs in the past in other projects that considered
0 to be an EOF and this made me think there could be an issue here.

I cannot completely confirm that this fixed a particular issue. I know
that the reliability increased after I made several changes including
that one. I had many disconnections before. As it wasn't documented, I
though it could be an error and hence submitted the fix. I haven't had
any infinite loop issue and transferred ca 10000 files (from 20k to
10M) in my unit tests.

Have you had problem with i == 0 creating an infinite busy loop ?

One could always have a separate delay when the return is 0. I still
think that not allowing 0 is a potential issue.

We can either let this patch out for the moment until I or someone
else come with a test case, or put it in and see if the busy loop
problem appears. What do you prefer ?

J