On Sat, Jul 03, 2021 at 04:49:34PM +0200, Gerhard Sittig wrote: > On Thu, 2021-07-01 at 23:44 +0100, Martin Ling wrote: > > > > The current architecture of libsigrok makes this difficult at the > > moment. Please see the following bugzilla entry for some previous > > thoughts on this: > > https://sigrok.org/bugzilla/show_bug.cgi?id=1276 > > Don't think that this is related (the Bugzilla item that you > reference above is about unified or common SCPI device > identification, which is totally not the issue here).
It is very relevant in my view. >From the point of view of a Fluke 45 user, the potential pitfall of the local echo feature was previously detected and addressed by the driver. That detection had to be removed because it was being applied to other devices (#1272). That is a problem with the current architecture, in which every driver's scan routine is applied to every available device. The current approach means that changes to any SCPI driver can break other devices, in ways that an individual user/developer cannot reasonably be expected to test. I was responding specifically to this: > > On Thu, Jul 01, 2021 at 07:12:45PM +0100, John wrote: > > > Would be nice if we could have the best of all: other > > > devices unaffected,_and_ reliable operation for the fluke-45 > > > device,_and_ helpful diagnostics to users when a known fatal > > > condition is seen. Can we get there, please? In my view, achieving that goal in a maintainable way requires making the changes I proposed in #1276. Otherwise we will keep running into this sort of issue without realising it. If there was a central scan routine that got as far as an *IDN? response, then work on any driver could proceed from that point without any risk of breaking other hardware in the process. For any device that could not be reliably detected by the central routine, it would be clear manual selection is necessary. And if changes were needed to the central routine, then it would then be clear that these are risky and need extensive testing to be merged. Anyway, returning to the specific issue, I believe that a fully automated solution is possible for the Fluke 45. If local echo is enabled, this can be detected by looking for the echo of the *IDN? string itself. There is no need to send a separate string such as the "ECHO-TEST" used previously. If local echo is detected, we could then expect and discard the echoes of all subsequent commands. This would need support in the SCPI layer but in principle is straightforward. Then everything would just work, regardless of the echo setting. There would be no need to tell the user to change the setting, neither in an error nor in the wiki. But I am firmly of the view that this logic belongs in a central *IDN? probe routine which is tested on multiple devices. Regards, Martin _______________________________________________ sigrok-devel mailing list sigrok-devel@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/sigrok-devel
