[5.1] ABI/API checker: avoid printing fully qualified names in generic signature #26643
Conversation
Under ABI checking mode where we don't have sugar types, any printed name changes of type nodes worth raising an alert. rdar://45567621
…ature The heuristics to decide whether fully qualified type names should be printed may work differently when generating the baseline and when importing from just built frameworks. This patch makes it consistent so that such false positives won't happen. rdar://54276347
|
@swift-ci please test |
|
This seems backwards; you should prefer fully qualified names because otherwise you can get false negatives. |
|
In practice, other diagnostics will be emitted if a type has been moved around, e.g. "struct S has been removed". We shouldn't rely on generic signature changes to tell us about type changes. |
|
I'm not worried about changes to the type; I'm worried about getting, e.g., |
|
@jrose-apple ah, I see. Thank you for taking a look! I agree that fully qualified type names in the generic signatures seem to be a better choice in that case. Since we need this current PR as a minimum change to eliminate the false positives some framework authors saw, I plan to merge it as it is and update to using the fully qualified name on master (here: #26642 ). |
|
I guess either way we're breaking the baseline, so 👍 |
The heuristics to decide whether fully qualified type names should be printed may
work differently when generating the baseline and when importing from just built
frameworks. This patch makes it consistent so that such false positives won't happen.
rdar://54276347