Fix workflows for new mbedtls

[peterharperuk]

Anything that builds picotool now has to pass "recursive" flag to the "git submodule init" command

[will-v-pi]

This shouldn't be needed once raspberrypi/picotool#226 is merged

[peterharperuk]

I was wondering if we should go the other way and make more use of the mbedtls cmake files, although I haven't investigated if this is possible.

[will-v-pi]

One benefit of moving to the SDK style is it reduces the number of files picotool compiles, as it doesn't need the x509 or TLS files so can skip those lists, whereas I'm not sure if there's a way to do that when using the mbedtls cmake files?

But the disadvantage is obviously maintenance, as we need to update the source lists if the sources change

[peterharperuk]

reduces the number of files picotool compiles

Yes - that's a good reason!

[will-v-pi]

Probably good to merge this anyway

[lurch]

pass "recursive" flag to the "git submodule init" command

I'm slightly wary of doing that (and yes I'm aware that you've already done that in some of the recent pico-sdk PRs, e.g. raspberrypi/pico-sdk#2416 ) because I'm worried about a situation like we previously had with tinyUSB cropping up again (where a 3rd-party repo had a lot of unnecessary submodules that we didn't need, which made the recursive git clone really slow). It's obviously not directly relevant to pico-sdk / pico-examples, but for an example of a repo with a LOT of submodules, see e.g. https://github.com/micropython/micropython/tree/master/lib

IMHO it's better to explicitly only init the submodules that we actually need, rather than recursively init-ing ALL submodules. However I'm happy to adopt the "let's leave it for now, and cross that bridge if we get to it" approach 🙂 🤞

[kilograham]

pass "recursive" flag to the "git submodule init" command

I'm slightly wary of doing that (and yes I'm aware that you've already done that in some of the recent pico-sdk PRs, e.g. raspberrypi/pico-sdk#2416 ) because I'm worried about a situation like we previously had with tinyUSB cropping up again (where a 3rd-party repo had a lot of unnecessary submodules that we didn't need, which made the recursive git clone really slow). It's obviously not directly relevant to pico-sdk / pico-examples, but for an example of a repo with a LOT of submodules, see e.g. https://github.com/micropython/micropython/tree/master/lib

IMHO it's better to explicitly only init the submodules that we actually need, rather than recursively init-ing ALL submodules. However I'm happy to adopt the "let's leave it for now, and cross that bridge if we get to it" approach 🙂 🤞

yes, recursive init is bad anywhere users might copy, which workflows aren't, but there they can be bad because it takes extra time - better to explicitly checkout