Conversation
|
Note that this matches the behavior before #249, because anyhow used to check the wrong env var for this. 1.0.60 is the only version of anyhow that ever respected RUSTC_WRAPPER. |
|
I don't actually know of a use of |
dtolnay
left a comment
There was a problem hiding this comment.
I would prefer to ignore only rust-analyzer's buggy RUSTC_WRAPPER specifically. Does it have some kind of recognizable name we could look for?
I'll have to forward that question to @lnicola |
RA sets itself as the wrapper, so it should be called
|
|
This should help for RA, not sure if we want to start carrying a list of all the broken wrappers out there... |
|
I've found one use of |
|
OTOH, given that autocfg is widely used and cuviper/autocfg#26 has seen very little activity, it doesn't seem to be terribly common for RUSTC_WRAPPER to be used in a way that just ignoring it causes a lot of trouble. |
dtolnay
left a comment
There was a problem hiding this comment.
I put up a different workaround in #252 since I think I would prefer not to sweep this under the rug yet as done by this PR.
I would be open to the workaround from this PR only if rust-lang/rust-analyzer#12973 is accepted as a rust-analyzer defect and we see progress toward resolving it, or if someone makes a compelling case that rust-lang/rust-analyzer#12973 is not a rust-analyzer defect and rust-analyzer will therefore not change behavior. The thing I don't want is for this workaround to land and then no follow-up ever occurs about the root cause.
|
Sure, that also works. I locally disabled the rustc-wrapper hack anyway, just posted this because I felt bad for my previous PR triggering that problem. |
Works around rust-lang/rust-analyzer#12973 by un-doing parts of what #157 intended to do, but only #249 actually did: honoring RUSTC_WRAPPER.
@jonhoo are you aware of any cases where ignoring RUSTC_WRAPPER actually causes problems? Seeing as autocfg ignores RUSTC_WRAPPER, it seems unlikely to me.