[Ovmsdev] OBD poller module shell proposal
Michael Geddes
frog at bunyip.wheelycreek.net
Sat Mar 30 08:39:43 HKT 2024
Thanks so much for giving this a good run.
I totally missed that about the purpose of PollerStateTicker() damn.
I think I can call this on primary ticks only from PollerSend() itself
outside of the mutex. That would work - I'll get onto that.
In the original PR# 966 PollerSend() is an OvmsVehicle() member and as such
this is a no-brainer.
That would mean it will get called from the Rx/Poller task rather than the
schedule (ticker1) task which is not necessarily a bad thing... but
different.
I believe that for the Poller implementation though, this will need to be
an event similar to PollRunFinished() - which kind of has similar usage
but occurs between (effectively) poller ticks (ie when all the
entries in the list have been dealt with) rather than what will be at the
beginning of each Primary tick.
Unless you have an objection - I'll pass in (canbus* bus) to this - similar
to PollRunFinished()
//.ichael
On Fri, 29 Mar 2024 at 21:51, Michael Balzer <dexter at expeedo.de> wrote:
> Michael,
>
> first feedback: testing this revealed some strange issues with vehicle
> state changes apparently not being detected or reacted to properly.
>
> A quick first check shows you have changed the way, the
> PollerStateTicker() hook works: it's now called independently from / as of
> the component ticker.1 registration sequence after the poller's primary
> send, rather than before.
>
> The callback was explicitly meant to provide a means to change the poller
> state just before the next poll would take place:
>
> /**
> * PollerStateTicker: check for state changes (stub, override with vehicle
> implementation)
> * This is called by VehicleTicker1() just before the next PollerSend().
> * Implement your poller state transition logic in this method, so the
> changes
> * will get applied immediately.
> */
>
> This change is already present in PR #966, so I'll delay merging that.
>
> Apart from that, the new poller seems to work normally, i.e. the polling
> scheme works and results are coming in, at least regarding ISOTP. I haven't
> tested VWTP yet.
>
> Regards,
> Michael
>
>
> Am 26.03.24 um 08:46 schrieb Michael Geddes:
>
> Yeah wow. So close and yet so far.
>
> Fix pushed.
>
> //.
>
> On Sat, 23 Mar 2024 at 21:58, Michael Balzer <dexter at expeedo.de> wrote:
>
>> Michael,
>>
>> looking forward to test this, but trying to run your branch with the VW
>> e-Up leads to an early crash boot loop.
>>
>> Apparently `OvmsPollers::PollSetPidList()` doesn't check for a NULL plist:
>>
>> 0x40195b6f is in OvmsPollers::PollSetPidList(canbus*,
>> OvmsPoller::poll_pid_t const*, OvmsPoller::VehicleSignal*)
>> (/home/balzer/esp/Open-Vehicle-Monitoring-System-3/vehicle/OVMS.V3/components/poller/src/vehicle_poller.cpp:1201).
>> 1196 bool hasbus[1+VEHICLE_MAXBUSSES];
>> 1197 for (int i = 0 ; i <= VEHICLE_MAXBUSSES; ++i)
>> 1198 hasbus[i] = false;
>> 1199
>> 1200 // Check for an Empty list.
>> *1201 if (plist->txmoduleid == 0)*
>> 1202 {
>> 1203 plist = nullptr;
>> 1204 ESP_LOGD(TAG, "PollSetPidList - Setting Empty List");
>> 1205 }
>> 0x401a2675 is in OvmsVehicle::PollSetPidList(canbus*,
>> OvmsPoller::poll_pid_t const*)
>> (/home/balzer/esp/Open-Vehicle-Monitoring-System-3/vehicle/OVMS.V3/components/vehicle/vehicle.cpp:2278).
>> 2273 return m_pollsignal;
>> 2274 }
>> 2275 void OvmsVehicle::PollSetPidList(canbus* bus, const
>> OvmsPoller::poll_pid_t* plist)
>> 2276 {
>> 2277 m_poll_bus_default = bus;
>> 2278 MyPollers.PollSetPidList(bus, plist, GetPollerSignal());
>> 2279 }
>> 2280 #endif
>> 2281
>> 2282 /**
>> 0x4023209e is in OvmsVehicleVWeUp::OBDInit()
>> (/home/balzer/esp/Open-Vehicle-Monitoring-System-3/vehicle/OVMS.V3/components/vehicle_vweup/src/vweup_obd.cpp:233).
>> 228 obd_state_t previous_state = m_obd_state;
>> 229 m_obd_state = OBDS_Config;
>> 230
>> 231 if (previous_state != OBDS_Pause)
>> 232 {
>> *233 PollSetPidList(m_can1, NULL);*
>> 234 PollSetThrottling(0);
>> 235 PollSetResponseSeparationTime(1);
>> 236
>> 237 if (StandardMetrics.ms_v_charge_inprogress->AsBool())
>>
>>
>> Regards,
>> Michael
>>
>>
>> Am 21.03.24 um 10:17 schrieb Michael Geddes:
>>
>> I've pushed up my working tree here:
>> https://github.com/frogonwheels/Open-Vehicle-Monitoring-System-3/tree/new-poller
>> partly for feedback, and partly because I wanted it backed up!
>>
>> This incorporates about 10 pull/requests worth of work which is why I
>> didn't even add it as a draft, so I see this as taking some number of
>> months to push up.
>>
>> Of interest is really the final components/poller/src/vehicle_poller*
>> files and how that has all come together.. I've added some preliminary
>> documentation for that which I'm quite happy to accept any critiques or
>> requests for specific information!
>>
>> The Duktape metrics 'stale, age' methods are probably ready for a p/r
>> as-is.
>>
>> //.ichael
>>
>> On Sun, 17 Mar 2024 at 16:58, Michael Geddes <frog at bunyip.wheelycreek.net>
>> wrote:
>>
>>> Thanks Michael,
>>> I'll definitely add the config - pretty much sorted out the singleton
>>> (except for the change of directory). Just wondering if I should rebase
>>> this change down earlier or just keep it as the 'last change' that is
>>> making the final break. It might be a better transition that way?
>>>
>>> Anyway have a look at this little class below; we have a bunch of
>>> different implementations for an event register and came up with the
>>> OvmsCallBackRegister below.
>>>
>>> In my example the event notification looks like this:
>>> void PollRunFinished(canbus *bus)
>>> {
>>> m_runfinished_callback.Call(
>>> [bus](const std::string &name, const PollCallback &cb)
>>> {
>>> cb(bus, nullptr);
>>> });
>>> }
>>>
>>> I could possibly but it in main/ovms_utils.h ?? It could also use a
>>> std:map to implement it, though I think we save space this way and tbh the
>>> use case wouldn't be doing a lot of Register and Unregister which would be
>>> the slowest operations... I've made it not shrink the list size .... but
>>> again, not that important in the context.
>>>
>>> Thoughts?
>>>
>>> //.ichael
>>> --------------8< -----------------------------------
>>> /* Call-back register.
>>> * The list does not shrink which is fine for this use-case.
>>> * Can be made inexpensively threadsafe/re-entrant safe.
>>> */
>>> template <typename FN>
>>> class OvmsCallBackRegister
>>> {
>>> private:
>>> class CallbackEntry
>>> {
>>> public:
>>> CallbackEntry(const std::string &caller, FN callback)
>>> {
>>> m_name = caller;
>>> m_callback = callback;
>>> }
>>> ~CallbackEntry() {}
>>> public:
>>> std::string m_name;
>>> FN m_callback;
>>> };
>>> typedef std::forward_list<CallbackEntry> callbacklist_t;
>>> callbacklist_t m_list;
>>> public:
>>> ~OvmsCallBackRegister()
>>> {
>>> }
>>> void Register(const std::string &nametag, FN callback)
>>> {
>>> // Replace
>>> for (auto it = m_list.begin(); it != m_list.end(); ++it)
>>> {
>>> if ((*it).m_name == nametag)
>>> {
>>> (*it).m_callback = callback;
>>> return;
>>> }
>>> }
>>> if (!callback)
>>> return;
>>> for (auto it = m_list.begin(); it != m_list.end(); ++it)
>>> {
>>> if (!(*it).m_callback)
>>> {
>>> CallbackEntry &entry = *it;
>>> entry.m_name = nametag;
>>> entry.m_callback = callback;
>>> return;
>>> }
>>> }
>>> m_list.push_front(CallbackEntry(nametag, callback));
>>> }
>>> void Deregister(const std::string &nametag)
>>> {
>>> Register(nametag, nullptr);
>>> }
>>> typedef std::function<void (const std::string &nametag, FN
>>> callback)> visit_fn_t;
>>> void Call(visit_fn_t visit)
>>> {
>>> for (auto it = m_list.begin(); it != m_list.end(); ++it)
>>> {
>>> const CallbackEntry &entry = *it;
>>> if (entry.m_callback)
>>> visit(entry.m_name, entry.m_callback);
>>> }
>>> }
>>> };
>>>
>>> On Thu, 14 Mar 2024 at 12:08, Mark Webb-Johnson <mark at webb-johnson.net>
>>> wrote:
>>>
>>>> Michael,
>>>>
>>>> I suggest that if it is a separate component then better to move it to
>>>> it’s own component directory (just as canopen is done).
>>>>
>>>> For completeness, I suggest it would also be good to include
>>>> a CONFIG_OVMS_COMP_* sdkconfig (default: yes), and put that as a
>>>> requirement for your component (as well as for any vehicle doing polling, I
>>>> guess).
>>>>
>>>> Regards, Mark
>>>>
>>>> On 14 Mar 2024, at 11:01 AM, Michael Geddes <
>>>> frog at bunyip.wheelycreek.net> wrote:
>>>>
>>>> Thanks Michael, Mark,
>>>>
>>>> Sorry for not acknowledging earlier.. this feedback is great; I've just
>>>> been cogitating on the consequences.
>>>>
>>>> I still have the Poller hanging onto the vehicle by a thread so I
>>>> should just cut the thread making the Poller a separate singleton (it's
>>>> still embedded in the vehicle class for now with a small interface joining
>>>> them).
>>>>
>>>> If I do, does it need to get moved to a new directory or can it stay in
>>>> the vehicle/ directory? The file vehicle_poller.cpp (and the _isotp and
>>>> vwtp parts to it) are still pretty much as they were with only a change in
>>>> class name..
>>>>
>>>> I think I just need the poller to get its own values of m_can1 etc and
>>>> provide a different way of getting the feedback results.
>>>>
>>>> I also need to make sure I'm not cutting off the 'vehicle' class'
>>>> access to non-solicited messages (ie stuff that is just on the bus).
>>>>
>>>> //.ichael
>>>>
>>>>
>>>> On Mon, 11 Mar 2024 at 14:51, Michael Balzer <dexter at expeedo.de> wrote:
>>>>
>>>>> Actually, separating the poller from the vehicle was part of the plan
>>>>> of reworking it into a job/worker architecture. I see no reason the
>>>>> generalized poller would need to remain coupled to the vehicle.
>>>>>
>>>>> That's why I placed the OBD single request command in the "obdii"
>>>>> hierarchy (although a more proper naming would have been e.g. "isotp", but
>>>>> changing the name or having both would confuse users -- and meanwhile the
>>>>> poller also supports a non-ISO TP variant).
>>>>>
>>>>> Regards,
>>>>> Michael
>>>>>
>>>>>
>>>>> Am 11.03.24 um 00:51 schrieb Mark Webb-Johnson:
>>>>>
>>>>> Michael,
>>>>>
>>>>> It depends on whether the poller can *only* be used in the vehicle
>>>>> class or if it is a framework all by itself (for example with commands to
>>>>> manually poll specific PIDs, etc).
>>>>>
>>>>> If *only* within vehicle framework, then putting it as a sub-command
>>>>> under ‘vehicle’ seems sensible.
>>>>>
>>>>> If more general purpose, then perhaps look at ‘copen’
>>>>> (component/canopen) as an example.
>>>>>
>>>>> Regards, Mark.
>>>>>
>>>>> On 10 Mar 2024, at 7:25 AM, Michael Geddes
>>>>> <frog at bunyip.wheelycreek.net> <frog at bunyip.wheelycreek.net> wrote:
>>>>>
>>>>> Hi all,
>>>>>
>>>>> I know some of this (especially for the status) functionality is
>>>>> predicated on code that's not gone up yet - however this is allowing
>>>>> 'pause' and 'resume' of the poller (which has been merged).
>>>>> My question is not so much about the functionality and status
>>>>> information, but about the location of the *poller* subcommand. (See
>>>>> below).
>>>>>
>>>>> Should 'vehicle' be exclusively for switching the vehicle type?
>>>>> Should the 'poller' command be top-level? Under obdii?
>>>>>
>>>>> Thoughts welcome.
>>>>> If you do vehicle poller pause then the last line reads 'Vehicle
>>>>> OBD Polling is paused'
>>>>> //.
>>>>> -------8<----------------------------------------
>>>>>
>>>>> *OVMS# vehicle ?*
>>>>> Usage: vehicle [list|module|poller|status]
>>>>> list Show list of available vehicle modules
>>>>> module Set (or clear) vehicle module
>>>>> poller OBD polling status
>>>>> status Show vehicle module status
>>>>> *OVMS# vehicle poller ?*
>>>>> Usage: vehicle poller [pause|resume]
>>>>> pause Pause OBD Polling
>>>>> resume Resume OBD Polling
>>>>> *OVMS# vehicle poller*
>>>>> OBD Polling running on bus 1 with an active list
>>>>> Time between polling ticks is 1000ms with 1 secondary sub-ticks
>>>>> Last poll command received 1s (ticks) ago.
>>>>> Vehicle OBD Polling is running.
>>>>> _______________________________________________
>>>>> OvmsDev mailing list
>>>>> OvmsDev at lists.openvehicles.com
>>>>> http://lists.openvehicles.com/mailman/listinfo/ovmsdev
>>>>>
>>>>>
>>>>>
>>>>> _______________________________________________
>>>>> OvmsDev mailing listOvmsDev at lists.openvehicles.comhttp://lists.openvehicles.com/mailman/listinfo/ovmsdev
>>>>>
>>>>>
>>>>> --
>>>>> Michael Balzer * Helkenberger Weg 9 * D-58256 Ennepetal
>>>>> Fon 02333 / 833 5735 * Handy 0176 / 206 989 26
>>>>>
>>>>> _______________________________________________
>>>>> OvmsDev mailing list
>>>>> OvmsDev at lists.openvehicles.com
>>>>> http://lists.openvehicles.com/mailman/listinfo/ovmsdev
>>>>>
>>>> _______________________________________________
>>>> OvmsDev mailing list
>>>> OvmsDev at lists.openvehicles.com
>>>> http://lists.openvehicles.com/mailman/listinfo/ovmsdev
>>>>
>>>>
>>>> _______________________________________________
>>>> OvmsDev mailing list
>>>> OvmsDev at lists.openvehicles.com
>>>> http://lists.openvehicles.com/mailman/listinfo/ovmsdev
>>>>
>>>
>> _______________________________________________
>> OvmsDev mailing listOvmsDev at lists.openvehicles.comhttp://lists.openvehicles.com/mailman/listinfo/ovmsdev
>>
>>
>> --
>> Michael Balzer * Helkenberger Weg 9 * D-58256 Ennepetal
>> Fon 02333 / 833 5735 * Handy 0176 / 206 989 26
>>
>> _______________________________________________
>> OvmsDev mailing list
>> OvmsDev at lists.openvehicles.com
>> http://lists.openvehicles.com/mailman/listinfo/ovmsdev
>>
>
> _______________________________________________
> OvmsDev mailing listOvmsDev at lists.openvehicles.comhttp://lists.openvehicles.com/mailman/listinfo/ovmsdev
>
>
> --
> Michael Balzer * Helkenberger Weg 9 * D-58256 Ennepetal
> Fon 02333 / 833 5735 * Handy 0176 / 206 989 26
>
> _______________________________________________
> OvmsDev mailing list
> OvmsDev at lists.openvehicles.com
> http://lists.openvehicles.com/mailman/listinfo/ovmsdev
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.openvehicles.com/pipermail/ovmsdev/attachments/20240330/9a044a52/attachment-0001.htm>
More information about the OvmsDev
mailing list