Add note for identifier following by array literal error #108222
Conversation
|
r? @cjgillot (rustbot has picked a reviewer for you, use r? to override) |
rustbot
added
S-waiting-on-review
There was a problem hiding this comment.
I left some nits.
cc @pnkfelix who suggested this message -- To be quite honest, I'm not sure if the parser really benefits from such a bespoke recovery. I agree that the error message you originally hit was confusing, but this seems like it'll probably never get hit by someone else in production often -- at least with the way it's being implemented currently. Do you have thoughts? I don't want to block this parser change if you disagree.
obeis
force-pushed
the
array-literal-note
branch
from
48fdf43 to
93a3409
Compare
|
☔ The latest upstream changes (presumably #108488) made this pull request unmergeable. Please resolve the merge conflicts. |
|
r? @pnkfelix |
|
So, I first off want to apologize for how long its taken me to get around to this.
Looking at this PR, I'm inclined to agree with Michael's intuition above. This is a lot of dedicated code for a very narrow case. It is not the approach I expected to see here. I was expecting someting more along the lines of: First establish some kind of state saying "we currently think we are inside an index-expr (i.e. a pair of based on the rest of Michael's comment, its possible that I should accept that this issue just isn't going to be fixed at all. I'm not quite ready to give up to that level, but I am willing to say that we don't want a fix that looks like this PR. @obeis , if you're interested in working together to figure out what the right solution here looks like, please do ping me on Zulip (preferably in a fresh public topic under the #t-compiler stream) and we can try to hash this out. But in the meantime, I'm going to close this PR because I don't think it will land in this state. (Also, @obeis , thank you for your patience and for your contribution.) |
Closes #107518