Fix "run" button position in error index

[GuillaumeGomez]

This isn't really a rustdoc issue but I still made the same fix in the rustdoc.css file (doesn't hurt).

Before:

After:

cc @jyn514 (considering this is quite a big bug and an easy fix)
r? @Nemo157

[rust-highfive]

Some changes occurred in HTML/CSS/JS.

cc @GuillaumeGomez

[Nemo157]

@bors r+

[bors]

📌 Commit a4d7046 has been approved by Nemo157

[GuillaumeGomez]

@rust-lang/infra Once this fix is merged, can we update the current file for the error index for the current stable, beta and nightly please? (We can even update further back, the issue remains)

[jyn514]

This isn't really a rustdoc issue but I still made the same fix in the rustdoc.css file (doesn't hurt).

If it's not a rustdoc issue, then what is it?

[GuillaumeGomez]

As you can see, they don't use the rustdoc CSS on https://doc.rust-lang.org/nightly/error-index.html

[jyn514]

I don't understand. What bug is the rustdoc change fixing? Does it not use the rust.css file?

[GuillaumeGomez]

run.css is a short version of rustdoc.css, but it hasn't been updated in a long time. As for the change in rustdoc.css, it's mostly "just in case".

[GuillaumeGomez]

ping @rust-lang/infra Can update the rust.css on https://doc.rust-lang.org/error-index.html (for both stable and beta) please?

[Mark-Simulacrum]

I admit to being a bit confused by that - why would we do so out of band? If you want this on beta then a beta backport seems like the right strategy? (Likewise for stable though we're unlikely to have a point release this late in the cycle)

[GuillaumeGomez]

This is a weird situation considering that the css file I fixed is kinda "on its own".

[Mark-Simulacrum]

I don't understand. Is it not shipped with Rust? If it's not, then we need to update nightly too, right? I didn't think we had files not shipped with Rust - how does e.g. the error index work locally then?

[GuillaumeGomez]

I mean: the rust.css file used only for the error index doesn't follow the rustdoc update, so it might have been broken for quite a lot of time. So I guess that when we make a new release, this file is part of it, but considering how special it is, we need to "backport" this fix to stable and beta (of the error index only).

[Mark-Simulacrum]

So I guess that when we make a new release, this file is part of it

This tells me that... we need to beta-nominate this PR? (And potentially stable-nominate, but as I said, it's unlikely that there will be a stable point release this cycle).

If we're in agreement (as far as my reading of your comments goes) that the file would get updated with a beta backport, which seems definitively true - otherwise the error index wouldn't find it when viewed locally, which it does -- then the correct way to update it is to let this PR roll out as part of our release train. If we consider it a critical fix, then a beta backport seems like the right step, not an ad-hoc process.

You say "how special it is" but I admit to not seeing anything special about it - can you clarify? Why is this file special compared to all the other files? How does this impede using the standard process for beta and stable backports, but apparently not nightly?

[GuillaumeGomez]

I would have hoped to at least fix the current stable but beta is already that...

You say "how special it is" but I admit to not seeing anything special about it - can you clarify? Why is this file special compared to all the other files? How does this impede using the standard process for beta and stable backports, but apparently not nightly?

This file is special in the way that it's a shortened version of the rustdoc.css file. Meaning that even though it is generated using rustdoc, it doesn't fully use the rustdoc CSS changes because it uses its own CSS file (which is, in my opinion, not great). Like I said, the file is released the same as all the others, but I didn't expect that we didn't use the rustdoc.css file here, which is problematic because any change in rustdoc DOM might break the error index without us knowing about it. It's only by "luck" that I discovered this issue...

Anyway, I let you put the tags?

[Mark-Simulacrum]

So the file (rust.css) is not at all special with regards to distribution? Why did you ping the infra team specially then? If there's something that should be done in an ad-hoc way, then we should do it, but my current understanding is that even if I was to replace the files manually in the S3 bucket, it wouldn't actually help -- it would get replace by the standard deploy anyway.

[GuillaumeGomez]

It might be interesting to understand why the error index is using a special file though. This was for this reason that I pinged the infra team originally.

[Mark-Simulacrum]

Well, I don't think the infra team knows the answer to that question. Maybe git commit history can yield something if you looked; I'd guess it's just historical artifact.

[pnkfelix]

discussed at compiler team meeting. since this seems like its been broken like this for ages, and there is no ticket filed discussing the breakage, then it seems bad to beta backport a week before the release.

(Yes change seems clearly low risk, but approval sets a bad precedent given the (lack of) context.)

Declining to beta backport. (likewise for stable etc)