rogfer01 marked 15 inline comments as done.
rogfer01 added a comment.
I am retaking this, will upload an updated patch soon.
================
Comment at: lib/AST/Decl.cpp:3523
+ return !isBitField() &&
+ (this->hasAttr<PackedAttr>() || getParent()->hasAttr<PackedAttr>()) &&
+ Context.getDeclAlign(this) <
----------------
rsmith wrote:
> Does this properly handle anonymous struct/union members:
>
> struct A __attribute__((packed)) {
> char a;
> union { int b; };
> } a;
> int &r = a.b;
This test (modulo swapping `A` and `__attribute__((packed))` positions) works.
Do you think we may be missing some case here?
================
Comment at: test/SemaCXX/bind-packed-member.cpp:68-73
+struct F
+{
+ char c;
+ NonTrivialCopy x __attribute__((packed));
+ int w __attribute__((packed));
+} f;
----------------
rsmith wrote:
> Another interesting case:
>
> struct __attribute__((packed)) G {
> char c;
> NonTrivialCopy x;
> };
>
> Here, GCC ignores the `packed` attribute entirely (with a warning). For ABI
> compatibility, we should do the same.
Will do this in a later update of this patch.
I still have to check what is the best place to check this, though I think once
the class-specifier has been completed.
https://reviews.llvm.org/D23325
_______________________________________________
cfe-commits mailing list
cfe-commits@lists.llvm.org
http://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits
