ulab: actually update the submodule

[jepler]

#2802 missed the submodule update itself.

[jepler]

We may now face the first need to disable a ulab module, such as the newly added comparison module. At least the espruino pico board no longer builds with this PR, due to overflowing flash.

[v923z]

We may now face the first need to disable a ulab module, such as the newly added comparison module. At least the espruino pico board no longer builds with this PR, due to overflowing flash.

@jepler @tannewt I am willing to change the sub-module structure, if necessity be. I would, perhaps, suggest at this point to run a poll on which functions people need, and use the results as a guideline.

[jepler]

I chose to disable ulab altogether on the espruino pico.

[dhalbert]

Looks good to me. Removing on Espruino-pico is a fine choice.

[hierophect]

Looks good to me. Reminds me to re-check -flto, I can't remember why I didn't enable that last time but it may have been resolved by other recent makefile/linker changes.

[tannewt]

Are we able to disable submodules of ulab independently?

[jepler]

@tannewt yes, we are able to selectively disable parts of ulab. I originally investigated that as a solution, but with the new ulab.comparison module deactivated it left only about 2kB (I forget exactly) of flash free. My gut said, better to fully disable ulab on this board than return within days or weeks to revisit the issue. If we get lto and it reclaims 20+kB as I suspect it would, that would be a great time to re-revisit it. However, I can change this PR to disable as little as possible, if that's what you prefer. Please merge or let me know.

[tannewt]

This is fine. Thanks!

[v923z]

I chose to disable ulab altogether on the espruino pico.

This brings up another issue, which, perhaps, deserves its own thread: it would be really great, if users could customise their own ulab (or any other external module, for that matter). I have come to realise that in circuitpython the main tripping point is that the makefile, py.mk, has to be modified. Would it be possible to accept arbitrary make fragments from external modules, exactly as is done in micropython? With that, this issue of the espruino pico would be rendered redundant.

I believe, this would benefit everyone, because circuitpython users could then be able to pull directly from the ulab repository, and try whatever they want.

[tannewt]

@v923z We should discuss this on a separate issue. Generally, we don't support users integrating their own C code into CircuitPython.

[v923z]

@v923z We should discuss this on a separate issue.

OK. It is here: #2825

Generally, we don't support users integrating their own C code into CircuitPython.

I understand this, but here seems to be a legitimate reason for changing that.