The Race to Replace C and C++ • 2020 • Conference Guide • Handmade Seattle

0:43Abner Coimbre: Welcome to the podcast

1:12AC: What's your professional history with compilers, Andrew?

1:52Andrew Kelley: No professional history with compilers

2:49AK: Why Zig

3:37AC: How do people find Zig?

3:51AK: Zig's decentralised community¹

4:28AC: What's your relationship with LLVM?

4:37AK: Ever-loosening relationship with LLVM, as training wheels

5:42AC: So Zig began as a hobby and evolved into a full-time endeavour?

6:02AK: Zig's progress to a full-time project

6:29AC: Let's introduce Ginger Bill

7:06Ginger Bill: Hello!

7:09AC: What's your professional history of compilers, Ginger?

7:45GB: Zero professional history of compilers, and the evolution of Odin²^,3

10:02AC: This sounds like a not very academic endeavour

10:55GB: Practical learning

11:31AC: Was Zig also written with slight academic understanding?

11:48AK: Practical reasoning coupling with academic theory

12:49AC: We do have Josh

13:10Josh Huelsman: Hi!

13:15AC: Casual introduction of hidden complexity

15:34AK: Zig's "no hidden control flow, no hidden memory allocations"⁴

16:59AC: How does that compare to Odin?

17:04GB: Odin's context allocators

17:58GB: Simplicity is complicated

19:51AK: I love your point about "what does simple mean?"

19:58GB: Implicit worldviews

20:17AK: Information theory-based simplicity, file path encoding

21:19AC: Simplicity in Vulkan vs OpenGL

22:15AC: What's your take on simplicity, Josh?

23:13JH: Writing inline then splitting things off where this makes sense

23:56AC: But then Java wouldn't be the ideal language for you

24:00JH: You could technically do it

24:07AC: How does Odin handle encapsulation?

24:57GB: Odin's encapsulation, at the library level

25:58GB: Library structures: 1) C-like textual include

26:45GB: Library structures: 2) Python-like file-based

27:48GB: Library structures: 3) Modula-like directory-based

29:56AC: How does Zig handle interfacing with future code?

30:46AK: Languages want to talk to C, to call machine code

31:14AK: Zig's ABI: 1) Using Clang to parse C

32:08AK: Zig's ABI: 2) Exposing Clang in Zig to compile C

32:51AC: User-simplicity, thanks to developer-complexity

33:22AK: Zig development work for each LLVM release

33:44AC: What's your experience of porting compilers?

34:51JH: Porting compilers

35:20JH: GNU's Symbol Versioning⁵

35:49AK: glibc specifically?

35:51JH: Some other software on GNU also does this

36:00AK: What parts of the compiler depended on libraries using Symbol Versioning?

36:10JH: GCC / Clang's compiler directive and Jai's @ symbol versioning

36:36AK: So you're talking about versioning the programs generated with the compiler?

36:46JH: How do Odin and Zig handle versioning?

37:12AK: Zig's Linux port: 1) Not linking glibc, using the stable syscall ABI

37:56AK: Zig's Linux port: 2) Executable generation, targeting a specific version of glibc, and static linking

38:53JH: Obviously you can't statically link OpenGL?

39:00AK: Yeah

39:01JH: How do you work with that?

39:06AK: Moby-Dick problem: Producing static binaries for games

39:27AC: Kid gloves in languages

40:50GB: "With great power comes great responsibility"

42:03GB: A brief history of programming languages

43:45GB: C's breakthroughs: 1) Type-system

44:18GB: C's breakthroughs: 2) Structured

44:54AC: Hardware-aware programming

46:10GB: "Programming is a tool to solve problems that you have in the domain of computers"

46:50GB: "The purpose of a program is, and ought to be, something that transforms data into other forms of data"

47:18AC: Josh and Andrew, are you vibing with us here?

47:25AK: Zig is a general-purpose language

48:45AC: Josh, didn't you write a brainfuck compiler recently?

48:48JH: brainfuck compiler

49:00GB: That sounds wonderful

49:03AC: How many years have you been a hobby compiler person?

49:10JH: Six years as a hobby compiler person

49:15AC: Why write compilers for fun?

50:44JH: Getting into compilers after Jonathon Blow started talking about Jai

51:35AC: How meta do you get with bootstrapping / self-compilation?

52:39AK: Bootstrapping Zig

55:31AC: Do you get harassed with the question of Zig being self-hosted?

55:44AK: The self-hosting question comes with a lot of positive curiosity energy

55:53AC: What about Odin in terms of usability?

56:21GB: Odin's usability, writing in it at JangaFX on the EmberGen project⁶

59:39AC: Let's dive into that "joy of programming"

1:00:20AK: Zig doesn't really prioritise the day-to-day experience of the programmer

1:00:29GB: Zig's maximisation of explicitness vs Odin's minimisation of implicitness

1:02:30AC: We're gonna want to dive into those philosophical differences

1:02:32AK: Zig optimises for the end-user of the software

1:03:11AC: Developer eating of complexity

1:03:27GB: Odin's approach to complexity, and recommending languages

1:05:13AK: Shots fired

1:05:15GB: I love Go

1:05:21AC: Gunshots

1:05:27GB: Rust's ever-complex vs Go's closely-defined approach

1:06:11AK: Go almost added Zig's "try"

1:06:15GB: I was hoping they didn't

1:06:20AK: I was thinking if they did, then Zig would be less far advanced than Go

1:06:32GB: Zig's exception-like error handling vs Odin's lack thereof

1:07:44AC: What's your approach to ABI compatibility?

1:09:31JH: You should not break the C ABI because everyone uses it

1:09:40AC: So why is Go doing it?

1:09:46GB: Google's self-hosting Go vs gccgo

1:10:12AK: Sounds like an abusive relationship

1:10:14GB: Go and C's one-way ABI compatibility relationship

1:11:07AK: And the C ABI is not really well-defined

1:11:20AC: Isn't there a System V ABI spec somewhere?⁷

1:11:24GB: System V vs MSVC ABI

1:12:01AC: I'm scared of breaking ABI compatibility

1:12:23GB: Language–interface separation

1:13:07AK: Compiler-decided internal ABI and external calling conventions

1:13:25JH: That's how it was in Jai as well

1:13:43AC: Improving C's ergonomics

1:15:39AK: You can enable a switch for implicit switch-statement fall-through

1:15:53AC: Are Zig's ergonomics informed by C's?

1:16:40AK: The Road to Zig 1.0⁸

1:17:07AC: How about Odin's ergonomics in relation to C?

1:17:18GB: Everything you wanted in C I've added to Odin

1:19:09AK: How do you get the Unicode data for the string stuff?

1:19:21GB: Odin assumes all strings are UTF-8 and has a string type

1:20:14AK: If the user does toupper(), where are you getting that data from?

1:20:22GB: Unicode table look-up⁹, Odin's rune codepoint type and C's three byte-types

1:21:17AK: You're talking about char, unsigned char and signed char, right?

1:21:22GB: Correct, yes

1:21:23AC: That's definitely tripped me up in the past

1:21:26GB: Finding the semantics of C's byte-types

1:21:49AC: Josh, do you have other dream improvements to C's ergonomics?

1:22:29JH: I have less ergonomic complaints of C

1:22:52AC: I guess your focus is on a small set of language features?

1:23:13JH: Special- and general-purpose languages

1:24:20AC: The value and fun of language exploration

1:25:11AK: Undefined variables and semantic communication of values to our tools

1:27:42AC: What would you recommend for pointers assigned later than their initialisation?

1:28:07AK: Try to give it the value as soon as possible, but don't give it 0

1:28:17GB: Zero Is Initialisation (ZII): Making the 0 value useful by default

1:29:29AC: So passing a cleared struct to a library should elicit some correct (default) behaviour?

1:30:02GB: Yes

1:30:05AK: So does this mean that all the pointers or references in Odin can be null?

1:30:08GB: No! Pointer usage in Odin

1:31:20AK: Can pointers be null?

1:31:22GB: Pointers can be null, and handling pointer errors

1:32:26AK: Null pointers and (un)definition of variables are related

Conference Guide

Keyboard Navigation

Global Keys

Menu toggling

In-Menu and Index Controls

Quotes and References Menus and Index

Quotes, References and Credits Menus

Filter Menu

Filter and Link Menus

Credits Menu