rustdoc: escape the deprecated and unstable reason text

[estebank]

Fix #38220.
Instead of the current output:

display:

[rust-highfive]

r? @steveklabnik

(rust_highfive has picked a reviewer for you, use r? to override)

[petrochenkov]

@bors r+ rollup

[bors]

📌 Commit 4cfc1e3 has been approved by petrochenkov

[ollie27]

These strings are run through a Markdown renderer after this and it doesn't make any sense to HTML escape something and then treat it as Markdown. I think the right thing to do here is replace the Markdown rendering with just escaping.

Also there is one other instance of this below these two that should probably be changed as well.

[estebank]

@petrochenkov travis error seems unrelated

[petrochenkov]

@bors r-
@bors delegate=ollie27

[bors]

✌️ @ollie27 can now approve this pull request

[ollie27]

Actually we could leave it using Markdown and just add `s around Box<FnOnce> in the stability message like is done elsewhere.

[estebank]

@ollie27 I'm thinking that even if using Markdown syntax to mark code would be preferable here it should still look reasonable (as in escaped) even if the code author forgets to do so.

HTML Escaping the text of the unstable reasons and then passing Markdown syntax will only avoid people from using HTML anchors for links or HTML tags for formatting, little more. I don't think any part of the markdown syntax would be affected by the escaping in the first place :)

[ollie27]

Adding the quotes and escaping results in it being rendered as "will be deprecated if and when Box<FnOnce> becomes usable", so yes escaping first does break Markdown.

[estebank]

@ollie27 you're absolutely right.

[steveklabnik]

/cc @rust-lang/tools , how do we want to move forward here?

[alexcrichton]

Sounds like @ollie27's suggestion may be the least invasive?

[estebank]

After reading through Hoedown's documentation, I have added a new MarkdownHtml struct that escapes HTML tags from the output, presenting the following output:

for reason "will be deprecated if and when `Box<FnOnce>` Box<FnOnce> becomes usable".

[estebank]

@ollie27 I feel that the patch is ready for merging. It changes this specific instance of the issue to use the code formatting, and avoids the issue going forward whenever somebody actually forgets to do so by escaping correctly by relying on Hoedown's feature flags. Only drawback is that you can't use HTML tags in these messages, but I feel that no one would be missing that feature for this field.

[ollie27]

I'm not sure about having the Markdown rendering act differently here to other places but as you've said it probably won't cause any problems in reality. We'll have to see what other people think.

Either way, if this change is accepted it will need tests.

[bors]

☔ The latest upstream changes (presumably #38329) made this pull request unmergeable. Please resolve the merge conflicts.

[steveklabnik]

ping @ollie27 are you happy with this now?

[ollie27]

Yeah this is fine with me. Are people okay with no longer allowing raw HTML in depreciation and stability messages?

[nrc]

@ollie27 sgtm

[estebank]

@ollie27 if you're happy with the PR, can you post "@bors r+" for @bors to automerge this change? :)

[ollie27]

Sure.

@bors r+

[bors]

📌 Commit e766c46 has been approved by ollie27

[bors]

⌛ Testing commit e766c46 with merge 9749df5...

[bors]

☀️ Test successful - status-appveyor, status-travis
Approved by: ollie27
Pushing 9749df5 to master...