Kill serde_codegen 💣

[nox]

This change is 

[jdm]

@bors-servo: r+

[highfive]

Heads up! This PR modifies the following files:

• @asajeffrey: components/constellation/Cargo.toml
• @KiChjang: components/net/Cargo.toml, components/script/Cargo.toml, components/net_traits/Cargo.toml, components/net_traits/Cargo.toml
• @fitzgen: components/script/Cargo.toml
• @emilio: components/layout/Cargo.toml

[highfive]

Warning

• These commits modify net, layout, gfx, and script code, but no tests are modified. Please consider adding a test!

[bors-servo]

📌 Commit f8467a2 has been approved by jdm

[bors-servo]

⌛ Testing commit f8467a2 with merge 74e7afa...

[bors-servo]

💔 Test failed - arm32

[nox]

Network issues, waiting for the other builds to finish.

[glennw]

@bors-servo retry

[bors-servo]

⚡ Previous build results for android, arm64, linux-dev, linux-rel-css, mac-dev-unit, mac-rel-css, mac-rel-wpt1, windows-gnu-dev, windows-msvc-dev are reusable. Rebuilding only arm32, linux-rel-wpt, mac-rel-wpt2...

[bors-servo]

☀️ Test successful - android, arm32, arm64, linux-dev, linux-rel-css, linux-rel-wpt, mac-dev-unit, mac-rel-css, mac-rel-wpt1, mac-rel-wpt2, windows-gnu-dev, windows-msvc-dev
Approved by: jdm
Pushing 74e7afa to master...

[edmorley]

As a head's up, the commit message here contained a non-BMP unicode character ("astral unicode character"), which requires 4 bytes to represent rather than the 3 bytes of BMP unicode, and as such causes issues with MySQL tables using utf8. This is affecting Treeherder pushlog ingestion (https://bugzilla.mozilla.org/show_bug.cgi?id=1115608) - we'll either need to strip out the non-BMP unicode or else switch this particular table to utf8mb4.

Obviously Treeherder should definitely handle this case gracefully, but there are may be other systems that consume the hg.mozilla.org pushlog that may not handle it well either, so please bear it in mind when using more unusual unicode characters in commit messages :-)

[nox]

@edmorley How was it not a problem in the past though? This is not the first time I use emojis.

[edmorley]

I imagine they were within the BMP (Basic Multilingual Plane) unicode range:
https://en.wikipedia.org/wiki/Plane_(Unicode)

(or else only in the PR description and not the commit message)

[nox]

I'm pretty sure U+1F916 ROBOT FACE is outside the BMP.

[nox]

Oh right, you squash all commits of a PR together, right?

[SimonSapin]

we'll either need to […] switch this particular table to utf8mb4.

Yes. It’s been in MySQL since 2010.

I personally don’t think that “Only use BMP characters” is much more acceptable than “Only use ASCII characters”.

As to PR descriptions, the get included by bors/homu in the message of merge commits.

[edmorley]

I agree, it's just unfortunate that:
(a) MySQL makes it way too easy to fall into this trap by defaulting to latin1 thus making people have to pick a charset themselves - which is easy to get wrong due to: their unfortunate naming of BMP-only utf8 as utf8, the sheer volume of docs online that don't even mention utf8mb4, and their not deprecating the BMP-only charsets. And after all that they then make people who want to convert utf8->utf8mb4 jump through hoops to support longer indexes (eg the innodb_file_format=barracuda, innodb_large_prefix=1 dance).
(b) Django doesn't support utf8mb4 out of the box (https://code.djangoproject.com/ticket/18392; not helped by MySQL being seen as a 2nd-class citizen by the Django project compared to PostgreSQL)

As another datapoint, bugzilla.mozilla.org is currently facing similar "utf8 doesn't mean utf8 in MySQL" realisations / hassle converting:
https://bugzilla.mozilla.org/show_bug.cgi?id=1253535
https://bugzilla.mozilla.org/show_bug.cgi?id=1328659

That said, I think the answer is just to give up on MySQL (this is just the last in a long list of annoyances we've had with it) and use PostgreSQL. But not an overnight switch for large existing projects like Treeherder. One day! :-)