thakis added inline comments.
================
Comment at: clang/test/CodeGenCXX/mangle-ms-auto-templates.cpp:17
+ AutoParmTemplate<0> auto_int;
+ // CHECK: call {{.*}} @"??0?$AutoParmTemplate@$H0A@@@QAE@XZ"
+ AutoParmTemplate<false> auto_bool;
----------------
zequanwu wrote:
> thakis wrote:
> > Are you sure this is correct? MSVC produces a different mangling
> > (https://godbolt.org/z/VxRfJd) and neither `undname` nor `llvm-undname` /
> > `demumble` can demangle the symbol here (while they demange the msvc output
> > according to godbolt fine).
> I use `x64 msvc v19.24` version, which gives
> `@"??0?$AutoParmTemplate@$MH0A@@@QAE@XZ"`. The extra `M` might come from
> qualifier mangling.
>
> For `x86 msvc v19.24(WINE)` version, it produces
> `??0?$AutoParmTemplate@$0A@@@QAE@XZ` for both `AutoParmTemplate<0> auto_int`
> and `AutoParmTemplate<false> auto_int`. Isn't this a bug?
The test at the top says -triple=i386. If you have x64 mangling results in
here, you should use a 64-bit triple (ie -tripe=x86_64-pc-windows). If the
mangling is structurally different for different bitnesses, we should probably
have tests for both.
For symbols that can be exported, we need to match msvc's mangling to be
abi-compatible, no matter if we consider the mangling a bug or not.
CHANGES SINCE LAST ACTION
https://reviews.llvm.org/D80409/new/
https://reviews.llvm.org/D80409
_______________________________________________
cfe-commits mailing list
[email protected]
https://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits
