Relevant micropython commits:
ad4656b861f94277bed9647ca176e662ce5119e3 all: Rename BYTES_PER_WORD to MP_BYTES_PER_OBJ_WORD.
9fef1c0bde2f9642d383bd56aa112447384a84ba py: Rename remaining object types to be of the form mp_type_xxx.
5fdf351178df9a18df624ae0f5947d8a5a6bce40 py/gc: Don't include mpconfig.h and misc.h in gc.h.
On a cleanly wiped device, storage must be unlocked in order to create
and retrieve the serial number. However, storage unlocking happens in
boot.py, which is imported _after_ usb.py
We must therefore sidestep asking for the serial number.
core: Remove dangling module decls
core: Use new Cargo feature resolver, use external MacOS debug info
core: Rust docs improvements
core: Upgrade bindgen
core: Add test target to Rust
ci: build rust sources
build(core): .ARM.exidx.text.__aeabi_ui2f in t1 firmware size
It's an unwind table for softfloat function inserted by rustc, probably
can be removed to save 8 bytes:
599c58db70/link.x.in (L175-L182)
scons: Remove dead code
core: Move Rust target to build/rust
core: Replace extern with a FFI version
core: Add some explanatory Rust comments
core: Use correct path for the Rust lib
core: Remove Buffer::as_mut()
Mutable buffer access needs MP_BUFFER_WRITE flag. TBD in the Protobuf PR.
core: Improve docs for micropython::Buffer
core: Minor Rust docs changes
core: Rewrite trezor_obj_get_ll_checked
core: Fix incorrect doc comment
core: Remove cc from deps
fixup! core: Rewrite trezor_obj_get_ll_checked
core: update safety comments
using define causes compiler to over-optimize the code in display_init_seq
leading to larger differences in the resulting binaries when the flag changes
we want to keep the difference to minimum (1 byte) and using const volatile
instead of define achieves that
This avoids problems with large timeouts causing the scheduler queue to
think the time counter has overflown, and ordering the autolock task before
immediate tasks.
The maximum reasonable time difference is 0x20000000, which in
microseconds is ~8 minutes, but in milliseconds a more reasonable ~6
days.
expose ff.c constants, raise them as arguments to FatFSError
introduce NotMounted and NoFilesystem as subclasses of FatFSError with
the appropriate error code set
- different approach to vector manipulation - more on the fly operations. Prepared for fully offloaded operations, BP on Trezor with constant memory.
- memory requirements reduced from (4MN + const) to (2MN + const)
- more raw methods to avoid unnecessary encoding/decoding
- chunking improved, chunk size set as a constant, changed from 64 to 32, missing pieces implemented to cover also BP 16
- proof_v8 support discontinued, old hardfork, not needed anymore
- get_exponent register clash fixed (for large vectors)
- reduced heap fragmentation by removing some temporary allocations
- hashing with len and offset to reduce heap fragmentation by creating a sliced arrays
- use to() wherever possible to avoid allocations and return of mutable private object
- global functions start with _ prefix, reduce import footprint
- use __slots__ in classes to minimize footprint
ff.c has a lazy-mounting feature, where any filesystem call will mount
the volume if it can. This messes with predictability of the mounted
state, so all (except mount/unmount/mkfs) Python functions will first
check if the fs is mounted.
Instead of having possibly multiple FatFS objects, each with its own
`fs` struct, there is one global static fs_instance. This is to match
the mode of operation of ff.c, which assumes a global list of mounts,
and all functions operate on the global based on path.
Methods of FatFS were converted to functions on the fatfs module.
fatfs.unmount() does not call ff.c's unmount, but simply invalidates
fs_instance. This is basically what ff.c would do, except without
messing with ff.c's global list of mounts.
