Get the light client to compile for iOS

[tomaka]

So that it can be used as a library by application, or as a background service (if that's possible on iOS).

[rphmeier]

cc #4842

[tomaka]

I could tackle the technical part of this, but I don't have any iPhone, Mac or Apple device in general.

[5chdn]

Let's get you some devices then. 💻

[ratake]

I got parity compiling for and running on iOS – it was fairly straightforward because iOS is very similar to macOS. The changes I had to make were:

• Disable USB hardware wallet support
  • Did this by commenting things out, not sure of the best location for the target/feature flag checks.
• Fix compiling C++ libraries on iOS rust-lang/cc-rs#285
• Update gcc-rs to cc-rs paritytech/rust-snappy#1
• Add an empty open(url: &str) function for target_os = "ios"
• Patch heapsize to work for non-jemalloc builds (i.e. to use system provided malloc_usable_size)
  • This is complicated to upstream because there is no way to detect at compile time which allocator is being used.

[ratake]

It would still need #4842 to be usable inside of an iOS app or extension (also light syncing needs some love in general).

[rphmeier]

@ratake nice work! it's true, we do need a library as well as better light sync. #8075 should address some of the syncing issues but beyond that we need stronger guarantees from the consensus layer.

could you submit a PR?

[tomaka]

I've looked a bit into this issue today, as I had exceptionally access to a Mac.

The heapsize situation is quite problematic.
We should be able to replace it with the malloc_size_of crate of servo, but that means we have to propagate the change throughout the dependency tree.
EDIT: malloc_size_of pulls up half of the Servo ecosystem.

[tomaka]

We also need cc-rs to pass a flag to the linker when building a library. Unfortunately, the rust-crypto crate still uses gcc-rs and doesn't seem to be maintained anymore. Hence the build fails.

Either we fork rust-crypto to use cc-rs, or we remove the rust-crypto dependency entirely. I would suggest the latter.

[adria0]

Closing issue due to its stale state.