Will the real sqUnixSocket.c step forward? (call for windows
tester)!
John M McIntosh
johnmci at smalltalkconsulting.com
Fri May 10 14:28:51 UTC 2002
>Nevin Pratt <nevin at smalltalkpro.com> wrote:
>> So far I haven't noticed any hiccups, on either FreeBSD (Intel) or
>> RedHat Linux. It seems to work fine.
>>
>> So, what advantage would taking the time to build again (thus using
>> Ian's updates) bring?
>>
> > Nevin
>
First we were able to run all the Socket Tests as outline on the
SUnits on my web site. In doing this resolved two problems.
1) A condition where in some instances asking if there was readable
data on a socket replied yes, but in fact there was no readable data.
This might result in a failure.
2) Simultaneous reads and writes on a socket using two processes
works as expected. Before Ian got the semaphore logic correct you
could experience 45 second pauses in traffic flow once you exceeded
the amount of tcp buffer space for a socket. I would be interested to
know if a similar problem exists on Windows but require someone that
has a real windows machine and a unix box to test against to run the
test suite to see if this problem exists on the windows
implementation.
http://www.smalltalkconsulting.com/html/OTNotes4.html
I filed this code into a current 3.2 image and didn't have any
problems with running the sunits. I'd be delighted to work thru any
failure conditions with the tester... So we can have platform
interchangabilty on the socket prims.
Note the unix testing did point out one issue. On creating a udp
listening socket it's expected the address is 0.0.0.0, versus your
interface address. The mac code was hacked to provide the interface
address but I've been informed this is technically incorrect.
--
--
===========================================================================
John M. McIntosh <johnmci at smalltalkconsulting.com> 1-800-477-2659
Corporate Smalltalk Consulting Ltd. http://www.smalltalkconsulting.com
===========================================================================
More information about the Squeak-dev
mailing list
|