Pass the potential id to be used as the type wrapper id, and prevent leaving a dangling a type parsing template aliases. #105
Conversation
|
r? @nox |
|
No, it doesn't seem to fix that issue. |
|
Maybe the next one does, I'm trying to build a test-case now. |
emilio
force-pushed
the
msvc
branch
2 times, most recently
from
75388f5
to
587b54c
Compare
emilio
changed the title
|
☔ The latest upstream changes (presumably #110) made this pull request unmergeable. Please resolve the merge conflicts. |
|
Rebased (a while ago). r? @nox |
|
(btw, the second patch makes the MSVC panic go away) |
|
Actually, @fitzgen, may you also be interested in reviewing this? :) |
There was a problem hiding this comment.
In general, things make sense, as much as they can.
You mention in the title that this prevents a dangling type, but I don't see any new test case (hint hint).
This makes me think we should have a Trace trait for our IR types that gives all the ItemId a given IR thing has references to, and then we can walk the full IR and ensure that we don't have dangling ItemId references. Debug only, ofc.
This should fix the MSVC problems. Signed-off-by: Emilio Cobos Álvarez <ecoal95@gmail.com>
…emplate parameter case.
After the rewrite, this works. Fixes #3
|
Yep, added a test. The I also added a test for #3, which is r=me. @bors-servo: r=fitzgen,emilio |
|
📌 Commit e04eda1 has been approved by |
|
☀️ Test successful - status-travis |
This should fix the MSVC problems @upsuper has been having.
@upsuper, can you confirm if this fixes the panic you were hitting?