NEURON: Pointer to src is not in the currently accessed section <Astrocyte.Astrocyte object at 0x14da04650b20>.soma

[ikitayama]

    for i in range(ncell3):
      seg = settings.pc.gid2cell(i + rings[0].gid3start)
      settings.pc.source_var(ca[cyt].nodes[100]._ref_concentration, i + 2000, sec=seg.soma)

Is this supposed to work?

[nrnhines]

I don't see anything wrong with the syntax. seg is a cell. @ramcdougal is the RxD concentration properly formed? I see the documentation https://nrn.readthedocs.io/en/latest/python/modelspec/programmatic/network/parcon.html#ParallelContext.source_var
needs updating since source_var is no longer restricted to voltage in CoreNEURON as of Jan 1,2021 and in NEURON, sinc 2016

[ramcdougal]

So that's one way of specifying an rxd pointer, yes. I think the problem is that source_var's checking of sections doesn't know how to look up rxd things.

[ikitayama]

wondering if it is easy to fix.

…

On Wed, Nov 10, 2021 at 1:03 ramcdougal ***@***.***> wrote: So that's one way of specifying an rxd pointer, yes. I think the problem is that source_var's checking of sections doesn't know how to look up rxd things. — You are receiving this because you authored the thread. Reply to this email directly, view it on GitHub <#1523 (comment)>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/AAHLBZ5FB3MPMRYFP62MUFTULFA67ANCNFSM5HNASULQ> .

[nrnhines]

For the short term I can start a branch that uses the old pointer if it can't figure it out during update. Of course that relies on the rxd pointer being stable. That should be ok if all your rxd setup is finished prior to your calls to pc.source_var. Let me know if that workaround would be helpful.

[ikitayama]

Works for me.

…

On Wed, Nov 10, 2021 at 20:21 nrnhines ***@***.***> wrote: For the short term I can start a branch that uses the old pointer if it can't figure it out during update. Of course that relies on the rxd pointer being stable. That should be ok if all your rxd setup is finished prior to your calls to pc.source_var. Let me know if that workaround would be helpful. — You are receiving this because you authored the thread. Reply to this email directly, view it on GitHub <#1523 (comment)>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/AAHLBZ4ZHIQDM5JFOQILX73ULJIVNANCNFSM5HNASULQ> .

[nrnhines]

Does the error message you receive look like Pointer to src is not in the currently accessed section <name>

[nrnhines]

Turned out the effort was more than I self-bargained for but I think the https://github.com/neuronsimulator/nrn/tree/rxd-sourcevar
will allow you to use an rxd variable reference as a source_var. This is likely to be the basis for a more complete implementation that is able to update these kinds of pointers. Please note the simple test in nrn/test/pynrn/test_partrans.py in the neighborhood of line 264 where I play a ramp into the hoc variable partrans_foo and that gets copied every time step to an
IClamp.amp. Also note in that file that I had to comment out 3 expect_error tests having to do with the normal requirement that a source variable be associated with a Section so that the pointers can be updated if necessary. I strongly recommend that you instrument some test code in your model that tests that after finitialize and after psolve that all source_var are equal to target_var. Or at least fill all the sources with values that identify them in some way (eg relate to the sid) and verify after an
finitialize or fadvance, that those values are present in the targets with the same sid.

[ikitayama]

NEURON: Pointer to src is not in the currently accessed section <Astrocyte.Astrocyte object at 0x1459696a0d60>.soma
 near line 0
 ^
        ParallelContext[0].source_var(..., 2000)
Traceback (most recent call last):
  File "ringtest.py", line 234, in <module>
    settings.pc.source_var(ca[cyt].nodes[100]._ref_concentration, i + 2000, sec=seg.soma)
RuntimeError: hocobj_call error

Is the above expected error (I'm using @nrnhines 's dev branch)?

[ikitayama]

Cc: @tatenokatsumi

[nrnhines]

Are you sure you are using the branch. That error message was commented out and replaced by a Notice.

-       hoc_execerr_ext("Pointer to src is not in the currently accessed section %s", secname(sec)); return NULL; //avoid coverage false negative.
+//     hoc_execerr_ext("Pointer to src is not in the currently accessed section %s", secname(sec)); return NULL; //avoid coverage false negative.
+       Fprintf(stderr, "Notice: ssid=%ld assumes fixed pointer %p (present value %g)\n", (long)ssid, pv, *pv);
+       return NULL;

Feel free to comment out the notice if you see too many of them.

[ikitayama]

Ah, you are right.

$ python ringtest.py
0.036
in define_biophycis
0.0800002s created rings
Notice: ssid=2000 assumes fixed pointer 0x2cae2d0 (present value 0.0001)
Traceback (most recent call last):
  File "ringtest.py", line 238, in <module>
    syn = target.synlist3[i]
IndexError: list index out of range

Is this the expected error?

[nrnhines]

The Notice makes sense given this issue.. The IndexError is a python error that is likely a bug in your program.

[ikitayama]

It seems now working. Thanks.

[ikitayama]

@nrnhines is this merged upstream?

[ikitayama]

I'd like to switch back to master if that's the case.

[nrnhines]

@pramod @ramcdougal This was not made a pull request. For the short term I can (will) make this a pull request, mark it as "work in progress" and merge the master into the branch so it will be up to date. The assumption of rxd pointer stability is not something I'm sure is warranted. And I haven't discussed this in detail with Robert. At any rate, turning off the code that raises an error if a pointer can't be updated is too generic, even though it's ok for your rxd pointers under your usage circumstances.
I (and I hope Robert can participate in the design) have to come up with a way to check for rxd pointer validity/updating.