Fix main.js for gtk-rs

[GuillaumeGomez]

We need it for gtk-rs. :)

r? @rust-lang/rustdoc

[jonas-schievink]

Can you add a comment explaining what it's doing and why?

[QuietMisdreavus]

What is gtk-rs doing that requires this fix?

[GuillaumeGomez]

It's checking that the values exist before using them.

[jonas-schievink]

Is that only needed by gtk-rs, or would all crates benefit? Why or why not?

[GuillaumeGomez]

I assume that if we had this issue with gtk-rs, other crates might encounter it as well.

[jonas-schievink]

What issue though? Is there an upstream issue you can link to? Why is gtk-rs hitting this and (apparently) no other crates?

[GuillaumeGomez]

I think it's because we're generating a lot of crates at once, which is rarely done outside (or just locally).

Another difference is that we generate without crates' docs without minification and then run the minification. However, the process is the same as in rustdoc so chances that it comes from there are quite small.

[Dylan-DPC-zz]

r? @GuillaumeGomez

[Dylan-DPC-zz]

oops
r? @QuietMisdreavus

[QuietMisdreavus]

I think it's because we're generating a lot of crates at once, which is rarely done outside (or just locally).

What do you mean by "a lot of crates at once"? That sounds to me like a regular cargo doc run of any crate with a deep dependency tree, or a workspace with many crates in it (like the compiler).

Another difference is that we generate without crates' docs without minification and then run the minification. However, the process is the same as in rustdoc so chances that it comes from there are quite small.

In other words, you run your crates' docs with --disable-minification except for the last one? (Or do you run some client of the minifier crate over it after all the crates have been run?) That sounds like a minifier bug more than a rustdoc bug.

---

To be clear, the actual change in this PR is small and not even that controversial. What i'm concerned about is that it looked like there was no investigation into what was going on, and that it was framed as a fix specifically for gtk-rs. It sets a weird precedent if one of us can shotgun our fixes through without so much as opening an issue (which you usually do for your small changes anyway, but not here) if the bug personally affects our own other projects, but anyone else with large/nonstandard project structures (e.g. rustc, Servo, Fuchsia) has to follow the usual process of opening an issue, contributing to an investigation, and waiting for a PR to open, review, and land.

[Dylan-DPC-zz]

ping from triage @QuietMisdreavus what's the status of this?

[QuietMisdreavus]

I never heard back from @GuillaumeGomez about what gtk-rs was doing that made this change necessary. I'm convinced this is papering up a different bug in the minifier crate, or in rustdoc's handling of it, instead of something the javascript itself would otherwise have to care about.

[GuillaumeGomez]

I wasn't able to determine if the bug was coming from rustdoc or my gtk-rs minification process... I can't use a workspace system so I'm stuck with using this fix for the moment.

[Mark-Simulacrum]

I'm closing this PR as the relevant procedure should likely be followed here; see #59958 (comment). Please reopen once that investigation is conducted and the PR description is updated.