Add the diagrams from my rustconf talk

[jyn514]

https://docs.google.com/presentation/d/1GnKlK98APzZ0Fw1BOqbroPQyvIio5Tm7SP7P0B8x3Cc/edit#slide=id.gf49216051b_0_96 (and the following 7 slides)

the sequence of slides, i.e. being shown multiple different diagrams in order, seemed to be particularly helpful - maybe we can find some way to do that with mdbook?

[eholk]

It'd be a pain to re-draw the diagrams, but you can do mermaid diagrams in mdbook. I don't know if that'd help make it so they can build on each other though.

[jyn514]

@RalfJung @ehuss @camelid you mentioned in #843 that you found my diagram confusing - do the diagrams here make more sense to you?

cc #914, @Mark-Simulacrum

[camelid]

Yeah, I think the new diagrams (from the talk) are clearer. I'm not sure exactly why, but I think the run-stage/build-stage distinction in the old diagrams was confusing. It's also helpful that the new one has version numbers and a graph representation rather than a table representation.

The one thing I found confusing in the new one is that in e.g. this diagram it says it's "Building" the beta compiler. I think it should say "Building with"?

[camelid]

I was able to adapt the diagram in mermaid, but I can't figure out how to get right angles and/or the nightly compiler and nightly std being at the same vertical position.

graph TD
    betac("beta compiler (1.63)") -->|A| std("nightly std (1.64)");
    betac -->|B| nightlyc{"nightly compiler<br>(1.64)"}
    std -->|"B (using 1.64)"| nightlyc
    nightlyc -->|C| std
    nightlyc -->|D| nightlyc2{"nightly compiler<br>(1.64)"}
    std -->|"D (using 1.64)"| nightlyc2

Loading

[jyn514]

I meant "building" to mean "the thing doing the building" - I agree "building with" is more clear, happy to change it.

[camelid]

One other thing I noticed is that "nightly compiler" has two separate nodes for each stage, but "nightly std" only has one. Shouldn't std be two as well?

[RalfJung]

I like it! It took me a bit to realize that the A, B, C on the arrows indicates the order though.

My main suggestion is that IMO it's confusing that the two standard libraries have the same node. It looks like there is a cyclic dependency when there isn't. Also you split up the two nightly compilers into two nodes, so for consistency IMO you should do the same with the standard libraries. That will also make the diagram more symmetric.

[camelid]

Here's the same diagram with two std nodes:

graph TD
    betac("beta compiler (1.63)") -->|A| std1("nightly std (1.64)");
    betac -->|B| nightlyc{"nightly compiler<br>(1.64)"}
    std1 -->|"B (using 1.64)"| nightlyc
    nightlyc -->|C| std2("nightly std (1.64)")
    nightlyc -->|D| nightlyc2{"nightly compiler<br>(1.64)"}
    std2 -->|"D (using 1.64)"| nightlyc2

Loading

[camelid]

Here's another option I made that I think is a bit clearer (though please let me know):

graph TD
    s0c["stage0 compiler (1.63)"] -->|A| s0l("stage0 std (1.64)");
    s0c & s0l --- stepb[ ]:::empty
    stepb -->|B| s1c["stage1 compiler (1.64)"]
    s1c -->|C| s1l("stage1 std (1.64)")
    s1c & s1l --- stepd[ ]:::empty
    stepd -->|D| s2c["stage2 compiler (1.64)"]

    classDef empty width:0px,height:0px;

Loading

[RalfJung]

Yes, that is the structure I was looking for. :) Not sure which of the two is better.

The remaining question is the labels... what you label "stage1 compiler" is the same as what bootstrap calls "stage 0 compiler artifacts", I think. So it might be worth having that as a separate node?

graph TD
    s0c["stage0 compiler (1.63)"] -->|A| s0l("stage0 std (1.64)");
    s0c & s0l --- stepb[ ]:::empty
    stepb -->|B| s0ca["stage0 compiler artifacts (1.64)"]
    s0ca -->|copy| s1c["stage1 compiler (1.64)"]
    s1c -->|C| s1l("stage1 std (1.64)")
    s1c & s1l --- stepd[ ]:::empty
    stepd -->|D| s1ca["stage1 compiler artifacts (1.64)"]
    s1ca -->|copy| s2c["stage2 compiler"]

    classDef empty width:0px,height:0px;

Loading

[camelid]

Ok, I've opened a PR!