Testing for the 4.4p1 release

Iain Morgan imorgan at nas.nasa.gov
Tue Sep 26 11:06:44 EST 2006


Sometime ago, Darren Tucker wrote:
> Iain Morgan wrote:
> > Sometime ago, Darren Tucker wrote:
> >> Could you please run "test/bntest" manually?  That will exercise the 
> >> code in my patch.
> > 
> > OK. Running test/bntest gives a lot of output. Based on the patch, I
> > assume you're only interested in the bn2dec test. That appears to be OK.
> 
> Alright, that passes the test cases I added for bn2dec (which admittedly 
> are a very small subset of the possible inputs it could encounter, and 
> much smaller than some of the inputs ssh would give it).
> 
> I'm tempted to generate a bunch of really big numbers at and test with 
> those too, but I don't much like the idea of nondeterministic regression 
> tests.
> 
> Based on the other failure you reported, I still suspect you have some 
> problem in OpenSSL although it's not exactly the same as the one I was 
> describing on HP-UX.  I found that using less aggressive compiler 
> options for OpenSSL stopped the problem from occurring, perhaps the same 
> would be the case here?
> 

Agreed. I don't recall the OpenSSL regression tests encountering errors
with previous versions that I've built, but I suppose I could have
overlooked them. Trying less aggressive optimization is certainly
worthwhile.

Thanks for the help.

--
Iain Morgan



More information about the openssh-unix-dev mailing list