FW: [globus-python] pyGlobus GSITCP raising exceptions instead of making me dig them out?
Robert Olson
olson at mcs.anl.gov
Tue Jul 13 20:09:46 CDT 2004
what's the official line on osx support, and has win support progressed?
On Tue, 13 Jul 2004, Ivan R. Judson wrote:
>
> So it would appear that the timeout issue will be something that *could* get
> fixed if we upgrade Globus, but we'd need windows support, and now OS X
> support to make that something we *can* upgrade.
>
> --Ivan
>
> -----Original Message-----
> From: Matthew Rodriguez DSD staff [mailto:MKRodriguez at lbl.gov]
> Sent: Tuesday, July 13, 2004 5:14 PM
> To: judson at mcs.anl.gov
> Cc: python-discuss at globus.org
> Subject: Re: [globus-python] pyGlobus GSITCP raising exceptions instead of
> making me dig them out?
>
> Ivan R. Judson wrote:
>
> >Is there any reason we can't have GSITCPSocket raise real exceptions
> >instead of making me use traceback.print_exc()? It feels really wrong
> >to be doing
> >this:
> >
> > try:
> > bytesRead = childSoc.read(buf, size, size)
> > except:
> > traceback.print_exc()
> >
> >When I could be doing something like this:
> >
> > try:
> > bytesRead = childSoc.read(buf, size, size)
> > except GlobusException, g:
> > log.exception("Error reading socket, details = %s",
> >g.globus_error)
> >
> >Implied in the example is the intention that GlobusException be a new
> >derived exception type, that has globus_error filled in with the result
> >of getting the last error from the globus layer.
> >
> >This would make a lot of programming easier.
> >
> >--Ivan
> >
> >PS -- We also want a timeout on the GSITCPSocket object, still.
> >
> >
> We'll fix the exceptions so that they have a globus_error member variable
> and they'll provide the stack trace as well. That's easy enough.
>
> Regarding setting a timeout on the GSITCPSocket, the globus_io interface
> does not provide a way to set a timeout. However the globus_xio interface
> does. So for versions of globus that have xio we could give the GSITCPSocket
> a settimeout method.
>
> Matt
>
>
>
More information about the ag-dev
mailing list