Pigweed Live Notes: 2023-08-28

[chadn...@google.com]

2023-08-28

Landed recently highlights (full changelog)

• pw_tokenizer
• Rust support! Docs
• Improved docs: API reference for tokenization & detokenization
• pw_toolchain: Major progress on Bazel toolchains
• pw_unit_test: Now returns structured data about the test run
• pw_web: Advanced filtering, performances fixes, and “jump-to-bottom” button to log viewer
• pw_presubmit: TypeScript linting and formatting
• pw_ide: --install-editable option for Python packages
• Docs: Parallel Sphinx builds
• Targets: Added Ambiq Apollo4, which uses the Ambiq Suite SDK and FreeRTOS

Coming soon

• Better Windows host support:
• Fixing bitrot, and making it much easier to set up Windows prerequisites
• Getting pw_system working on Windows
• Faster and easier on-boarding via IDE support
• Clone -> open in VS code -> start work!
• Focus is on VS Code now, but this will be easily accessible to other editors too
• pw_log support for Rust

Agenda

SEED-0109 review: Communication Buffers -

Taylor Cramer

• Rendered SEED Doc - CL
• Proposal: a standard buffer type for network-style communications
• If you have needs or expertise re: buffers, DMA, etc., check out the SEED and reach out

SEED-0108 review: Emulators Frontend

Octavian Purdila

• Rendered SEED Doc - CL
• Working with emulators in Pigweed isn’t easy right now, especially if you want to support multiple targets, platforms, projects, etc. You have to manage a bunch of stuff yourself!
• The SEED proposes an interface for specifying emulator configuration to take care of the above
• There are still some differences between host OSes and emulators, e.g., Windows doesn’t support PTYs, qemu and Renode have different device path specifiers – in other words, the emulator definition interface isn’t totally generic, it exposes some direct emulator configuration options where needed
• Easy to get into gdb! Next step: VS Code debugger integration

SEED-0105 review: Nested Tokens and Tokenized Log Arguments -

Gwyneth Chen

• Rendered SEED Doc - CL
• Runtime arguments to log statements aren’t currently tokenized – tokenizing these too can further reduce on-device log storage space
• Strings can be wrapped in PRI-style macros to ensure they’re tokenized
• This can be extended to enums too – wrap the enum definition in a macro to handle tokenization of the values
• Does this take up more space? 32-bit token for a < 32-bit enum value? No, the proposal is that the enum value is the token, so it doesn’t use the full 32 bits
• It’s about the same amount of data over the wire as transmitting the integer value, but is much more readable
• The SEED focuses on pw_log_tokenized, how does using the tokenization macros affect pw_log_basic (if at all)? Will address this in the SEED
• This contains the start of a Rust-like derive macro that could wrap strings and define printf-style encodings/formattings at compile time – super cool!
• Syntax considerations
• Developer syntax
• This is the main touchpoint for developers, so it should be easy and sensible – but it is changeable over time
• Wrapping enums can be tricky – and it’s awkward to have to specify the enum type at each log callsite. But this is still just a proposed syntax and there’s room for change
• Encoding/formatting syntax
• We’re committing to this! It will be embedded in token databases and implemented in multiple languages, so whatever we choose here, we will have to live with

Sent from document [PUBLIC] Pigweed Live Notes