Stubbing out the copy loop seems fine but what is the purpose of the example? It has no observable output and no mechanism to run it with the kind of bootloader that it must be used with to avoid UB. Was it LLM generated?

Hey @jonathanpallant the purpose of that example is to demonstrate how a Cortex-M program behaves when the usual .data section initialization is skipped by cortex-m-rt. This is useful because normally the cortex-m-rt copies .data section from flash to RAM. But for some scenarios when using the RP2040 boot2 bootloader as stated in issue #609 this copy has been already done by the bootloader itself, and the cortex-m-rt copies again, basically it reads from unmapped memory causing undefined behaviour.

Either way my PR isn't LLM generated, through that example I was indeed trying to show something reasonable :)

Thanks!

Ok so how do I run the example and what will I observe when the feature is set?

I think you might be able to just use a non-zero constant here to show that the data section was initialized? No need to use mutable statics then..

EDIT: the constant might be optimized away..

@KushalMeghani1644 have you actually tried if this solves the issue #609 is reporting, in actual RP2040 hardware, with the copy-to-ram boot2?

I'm closing this PR on the grounds it's taking far more effort to review than the change is worth.

I'm happy to look at this again when we have a working example (in another repo) with clear steps to reproduce the issue and clear evidence that the PR resolves the issue.