Dan Price wrote: > On Tue 15 Jul 2008 at 04:00PM, Tom Mueller (pkg-discuss) wrote: >> Please have a look at these changes for issue 993: >> >> issue 993: http://defect.opensolaris.org/bz/show_bug.cgi?id=993 >> >> webrev: http://cr.opensolaris.org/~tmueller/cr-993/ >> >> The unit test cases currently fail sometimes because the depotcontroller >> is able to open a TCP connection to the server and read a response >> before the server is really ready. By retrieving the catalog >> successfully, we make sure that the server is really up. >> >> Also, the server sometimes takes more than the allowed timeout to come >> up. I increased the timeout from 4 to 40 and these seems to help quite >> a bit with being able to reliably run the tests. > > I think the fix works as designed, but I am wondering if really this > represents a bug in the depot? It seems to me that the depot should not > start listening on port 80 prior to being truly ready to go.
Remember that with cherrypy, the depot is an application that runs on top of the http server. Thus while cherrypy may be ready, our application may not be. cherrypy does allow more fine-grained control over the startup process -- especially in the upcoming release. Until then, this might be a good solution. The other thing that comes to mind is the fact that the depot controller used by the test suite currently has to start an external program (depot.py) to do all of its work. I'm hoping to split of all of the depot functionality into a module in the near future so that the depot controller can simply a class that inherits from it and some of these issues might be further alleviated. -- Shawn Walker | Solaris Kernel Development | [EMAIL PROTECTED] | blogs.sun.com/srw _______________________________________________ pkg-discuss mailing list [email protected] http://mail.opensolaris.org/mailman/listinfo/pkg-discuss
