[const prop] Fix "alloc id without corresponding allocation" ICE

[wesleywiser]

r? @oli-obk

[rust-highfive]

Your PR failed (pretty log, raw log). Through arcane magic we have determined that the following fragments from the build log may contain information about the problem.

Click to expand the log.

2019-11-23T20:52:06.4600144Z ##[command]git remote add origin https://github.com/rust-lang/rust
2019-11-23T20:52:07.2108499Z ##[command]git config gc.auto 0
2019-11-23T20:52:07.2113036Z ##[command]git config --get-all http.https://github.com/rust-lang/rust.extraheader
2019-11-23T20:52:07.2114763Z ##[command]git config --get-all http.proxy
2019-11-23T20:52:07.2118784Z ##[command]git -c http.extraheader="AUTHORIZATION: basic ***" fetch --force --tags --prune --progress --no-recurse-submodules --depth=2 origin +refs/heads/*:refs/remotes/origin/* +refs/pull/66677/merge:refs/remotes/pull/66677/merge

I'm a bot! I can only do what humans tell me to, so if this was not helpful or you have suggestions for improvements, please ping or otherwise contact @TimNN. (Feature Requests)

[oli-obk]

Do we have any tests actually exercising the creation of a new const alloc and const propping a pointer to it without triggering promotion? I haven't looked into it yet, but I now also worry about interior mutability behind references.

[wesleywiser]

Do we have any tests actually exercising the creation of a new const alloc and const propping a pointer to it without triggering promotion?

I tried a couple different things and couldn't get that to happen. I think the blocker is this bit of code

rust/src/librustc_mir/transform/const_prop.rs

Line 618 in 388ffd9

this.ecx.try_read_immediate(value)

[oli-obk]

@bors r+

[bors]

📌 Commit 32e78ca has been approved by oli-obk

[bors]

⌛ Testing commit 32e78ca with merge 876a72a...

[bors]

☀️ Test successful - checks-azure
Approved by: oli-obk
Pushing 876a72a to master...