On Mon, Mar 29, 2010 at 11:24 AM, John Admanski <[email protected]>wrote:
> On Mon, Mar 29, 2010 at 11:22 AM, Darin Petkov <[email protected]> wrote: > >> >> >> On Mon, Mar 29, 2010 at 11:15 AM, John Admanski <[email protected]>wrote: >> >>> Right, but with this patch I would think that instead of prepending args >>> to the control file like in you original patch you could instead pass them >>> through via the new -a args? I don't know that it's any neater, from the >>> server point of view, but it at least passes them through the same codepath >>> in the client. >> >> >> I considered that too but there's some added complexity/fragility with >> going through shell args, python tuple, back to shell args... and escaping >> white-space, etc. This approach seemed simpler from that point of view. >> >> Darin >> >> > Yeah, I'm not sure the added complexity of translating it back into shell > args on the server side is worth the reduced complexity on the client of > only having one code path. This LGTM. > Thanks much for the prompt review and feedback. Martin, Please apply both patches when you get a chance. Thanks, Darin > > -- John > > >> >> >>> That's more of an implementation detail, though; I'm happy enough with >>> these two patches at least preserving feature parity between client-only and >>> client+server uses. >>> >>> -- John >>> >>> >>> On Mon, Mar 29, 2010 at 11:01 AM, Darin Petkov <[email protected]>wrote: >>> >>>> John, >>>> >>>> Attached is a separate patch that adds support for args to >>>> client/bin/autotest. The downside of doing it this way is that it >>>> implements >>>> a separate path for passing args to client-side control. The benefit is >>>> that >>>> it's simple. >>>> >>>> Darin >>>> >>>> >>>> On Mon, Mar 29, 2010 at 8:33 AM, John Admanski <[email protected]>wrote: >>>> >>>>> I was looking at this a bit more, and I still see an issue; now you >>>>> can't write code that uses args that just works on a straight client, you >>>>> have to be launching the client via autoserv. I suppose that's not a huge >>>>> problem, since autoserv is the only way that you can set args anyway, but >>>>> I >>>>> guess I'd just envisioned a grander patch that actually added support for >>>>> command-line args to client/bin/autotest, and then just used >>>>> server/autotest.py to pass through any autoserv args. >>>>> >>>>> Still, I think it's not a huge deal. I'm just a little worried about >>>>> the continued trend of more and more code being written that just assumes >>>>> you're using all the higher layers to run your tests. These days it seems >>>>> everyone just wants to write client tests that assume you're using >>>>> autoserv, >>>>> or even assumes you're using a full scheduler setup. >>>>> >>>>> -- John >>>>> >>>>> >>>>> On Fri, Mar 26, 2010 at 3:08 PM, Darin Petkov <[email protected]>wrote: >>>>> >>>>>> >>>>>> >>>>>> On Fri, Mar 26, 2010 at 2:43 PM, John Admanski >>>>>> <[email protected]>wrote: >>>>>> >>>>>>> This should probably use %r instead of %s. In practice I think >>>>>>> repr(args) and str(args) will end up being the same, but in principle >>>>>>> repr >>>>>>> is supposed to be the inverse of eval so it's the more correct choice of >>>>>>> formatting. >>>>>> >>>>>> >>>>>> Good point. Updated the patch. PTAL. >>>>>> >>>>>> >>>>>>> >>>>>>> >>>>>>> On Fri, Mar 26, 2010 at 2:40 PM, Darin Petkov <[email protected]>wrote: >>>>>>> >>>>>>>> A simple patch to propagate user args (autoserv -a/--args) to the >>>>>>>> client-side control file. >>>>>>>> >>>>>>>> >>>>>>> >>>>>> >>>>> >>>> >>> >> >
_______________________________________________ Autotest mailing list [email protected] http://test.kernel.org/cgi-bin/mailman/listinfo/autotest
