Kyle Hamilton wrote:
> Also, I assume it compiles with 0.9.7, 0.9.8, and 0.9.9?
I presume so, I'm working with 0.9.8b.
Also only tested on Linux. It should be simple to get running on any
unix with pthread support.
> How is it described -- a precompiled byte sequence?
Byte sequence ? Its a few lines of code to create and setup an
instruction and push it onto a linked list. There are opcodes for each
possible operation and a naive iteration scheme, which allows for
simulation of a 1 byte TCP window and retry of API call.
I need to think a little more on how I'd want to describe a testcase in
a text file. But its very easy to add new tests cases by adding a new
.c file and cloning test_basic.c and hooking it in.
Peter Sylvester wrote:
isn't ssl_test.c a sufficient starting point?
What is the purpose of it ? What is it testing. Maybe it needs
attention to get working again so we can run it. Including it in the
default built might stop it getting out of date with the rest of OpenSSL
project.
It looks like the first version of "openssl s_client" to me which is way
off the mark from where I'm at with sslregress.
My design would allow you to permutate every possible scenario of:
* Application high level calls to libssl.a
* Kernel buffering situation (read/write independently)
* Network data fragmentation / congestion (read/write independently)
That is if given enough test cases, or a test case generator program
spewed them out.
It would be possible to extend it for blocking socket simulation, random
testcase generation and handling multiple SSL connections between the
master and slave at the same time. These would be fairly straight
forward steps to take from here.
The code side is about ready, I'll just knock up a little documentation.
Darryl
______________________________________________________________________
OpenSSL Project http://www.openssl.org
Development Mailing List openssl-dev@openssl.org
Automated List Manager [EMAIL PROTECTED]