New bors looks within markdown code blocks

[jieyouxu]

It's very useful to be able to not trigger bors (old or new) when you want to show an bors invocation for whatever reason. In old bors, I believe there's special logic to prevent old bors to interpret @bors try or similar invocations within inline code blocks or code blocks. However, this doesn't seem to be carried forward to new bors.

Reproduction steps

• User writes

  Reminder (for myself): if try job fails, use @bors2 try to try out new bors

• Old bors: nothing
• New bors: tries to interpret as @bors2 try` to try out new bors and fails because ``try``` is not a valid command. See forward the bootstrap runner to run-make rust#141856 (comment).

Expected behavior

Ignored, because user is most likely trying to demonstrate the invocation and not actually run it.

Actual behavior (new bors)

New bors tries to run it.

[Urgau]

We have a module specifically dedicated to figuring out which part of the markdown we should ignore in triagebot. bors should follow it.

https://github.com/rust-lang/triagebot/blob/7044e5449aaf29eea939bc7db81c63b7c872b9b8/parser/src/ignore_block.rs#L10-L61

[jieyouxu]

Maybe worth pulling that out into some internal lib

[Kobzol]

I would love to have a shared monorepo for all our bots 😆 But that's probably not going to happen anytime soon. This is definitely a bug that should be fixed, but I wonder if we need such a heavyweight solution as triagebot uses.

I would be personally fine with saying that all bors commands have to start at the beginning of a line (modulo whitespace), otherwise they won't work. That would solve this issue on its own.

What do you think?

[Urgau]

Many people (including me) uses: r? Urgau @bors r+, would be good to not break it.

[Kobzol]

I see. Ok, in that case we'll need to shell out to a Markdown parser, I'll work on it.