Bug 49633 - Added a warning for global and namespace declared variables that are declared as both static and inline.
Details
Diff Detail
- Repository
- rG LLVM Github Monorepo
Event Timeline
clang/lib/Sema/SemaDecl.cpp | ||
---|---|---|
7127 | First, what about this is C++17 specific? The inline and static relationship is older than C++17, right? Also, are you sure that the warning/stuff in the 'else' isn't still valuable? |
clang/lib/Sema/SemaDecl.cpp | ||
---|---|---|
7127 | Perhaps I misunderstood something in the bug write-up, the component for the ticket is C++17. Also there is the warning that inline variables are a C++17 extension that appears when you use the inline keyword on a variable (regardless of the appearance of the static keyword). I suppose that does not necessarily mean we cannot simply show both warnings, and maybe that is the right thing to do. I felt that was not really necessary because of the other warning. As for the other warnings in the else the one is the warning that I mentioned (which only applies when the C++17 standard is not applied, and the other is a C++14 compatibility warning which states: You can see the messages for those diagnostic message here: Please let me know if I am still missing something with how this diagnostic was supposed to apply and where. Thank you. |
Should we also warn about inline variables in anonymous namespaces?
clang/lib/Sema/SemaDecl.cpp | ||
---|---|---|
7127 | This diagnostic should be independent of the ext / compat diagnostic below. This patch currently removes the -Wc++14-compat warning for static inline int n;, which would be a behavior regression for that warning. Moving this check inside the else block a few lines below, and removing the check for CPlusPlus17, would seem reasonable. We accept inline variables in earlier language modes as an extension, and we should still warn in that case. |
Updates to testing to ensure warning occurs for variables declared in anonymous namespace.
Updates per rsmith to ensure that ext compat warning still occurs fixing regressive behaviour.
First, what about this is C++17 specific? The inline and static relationship is older than C++17, right?
Also, are you sure that the warning/stuff in the 'else' isn't still valuable?