Fix issue 15309 again, without breaking the fix for regression 15550 #5390
Conversation
| { | ||
| if (ti.needsTypeInference(sc)) | ||
| { | ||
| // Bugzilla 15550: ti is a partial instantiattion form fun!tiargs. |
There was a problem hiding this comment.
Should be instantiation.
And would it be helpful to put some of this comment in the error message? Maybe expression (%s) is a partially instantiated template and has no type?
There was a problem hiding this comment.
Fixed typo,
The suggested message would be better, but diagnostic improvement is beyond the purpose of this PR.
I have a plan to move the "has no type" message into ScopeExp.checkValue().
There was a problem hiding this comment.
Ok. It would be nice to improve these errors as they can be unhelpful.
There was a problem hiding this comment.
By #5400, the error message for typeof(func!tiargs) will be improved.
Recognize partial instantiation form inside typeof()
|
Auto-merge toggled on |
|
Thanks! |
Fix issue 15309 again, without breaking the fix for regression 15550
| @@ -5332,6 +5332,9 @@ public: | |||
| return e.semantic(sc); | |||
| } | |||
| } | |||
| // ti is an instance which requires IFTI. | |||
| sds = ti; | |||
| type = Type.tvoid; | |||
There was a problem hiding this comment.
This remains still problematic, b/c whenever we don't handle ScopeExp correctly, a wrong type (void) might slip through.
How about using Type.terror instead? It should always be an error to use the type of an template instance that might need inference.
Issue 15309 has fixed in #5263, but it had introduced a regression issue 15550.
By @MartinNowak, 15550 has been fixed in #5368, but it has also reopened 15309.
Fix 15309 again.