use of settings consumes more flash (V2) - need to reflect this in available flash for app

[tballmsft]

• Context: the use of the settings feature (from pxt-common-packages) in pxt-jacdac reduces the amount of flash for the user code
• Problem: the web app and compiler are likely unaware of this, so may produce a binary (that uses settings) and overwrites the setting area in flash.
• Solution: need a way to adjust the amount of available flash in compile check for "user code too large", when settings is included in binary

[tballmsft]

There are target-specific settings for end of flash

            "compile": {
                "flashCodeAlign": 4096,
                "flashUsableEnd": 487424,
                "flashEnd": 524288
            },

There is no way to override on a package basis, so we would need some special case coding somewhere for this... or we just hard code the limit, assuming settings is used.

[tballmsft]

Note that the above flashUsableEnd already is incorrect, as it doesn't reflect the memory map: https://github.com/lancaster-university/codal-microbit-v2/blob/master/docs/MemoryMap.md.

From the memory map for the V2, the flashUsableEnd is now 0x73000, which is 471,040.

Subtracting 32k for settings get us down to 0x6b000 = 438,272

[tballmsft]

On further reflection, it seems rather severe to take away 32k for settings. On the other hand, most micro:bit programs are tiny! Thoughts?

[tballmsft]

@abchatra - Do we have a set of test programs we can benchmark on?

[tballmsft]

@carlosperate - can you please give me a set of the largest MakeCode programs for the micro:bit.

[tballmsft]

See #5548

[microbit-carlos]

• the use of the settings feature (from pxt-common-packages) in pxt-jacdac reduces the amount of flash for the user code

Is that the CODAL MicroBitStorage (or similar) settings.

If not, this could be also related:

• MakeCode programme can overlap reserved flash pages for BLE and CODAL storage #5403

@carlosperate - can you please give me a set of the largest MakeCode programs for the micro:bit.

I don't have anything specifically created for this purpose, mostly just encountered some examples (also from your own testing) like these:

• this program never works on V2 #5396
• MakeCode 2022 consumes more memory and some v4 programmes won't download anymore #4788

[abchatra]

I am assuming this doesn't need to be tracked as an issue. Closing. If not let me know.