[PATCH] Mangling of local names: anonymous unions, names in dead code

[Reid Kleckner]

>>! In D6295#5, @eastig wrote:
> You mean that this will hurt performance, don't you?

Yes, there is an (admittedly minor) performance cost to numbering static locals of non-inline functions when no numbering is required.

> The code that takes into account declarations in dead code was added in revision [[ http://llvm.org/viewvc/llvm-project?view=revision&revision=185986 | 185986 ]] by an engineer from Apple as a fix of the issue rdar://problem/14204721. I do not have access to Apple's bug tracking system to see what was the issue.

Yep, I'm familiar with the work Eli did here because I helped make the Microsoft ABI version of this work.

> The current issue is that different enumeration schemas are used by getNextDiscriminator. And they produce different names depending on the function specifier 'inline'.

They do. This is by design. Can you elaborate on why this is a problem? The names of static locals in non-inline functions are not ABI-relevant.

> According to the C++ standard [dcl.fct.spec]: A static local variable in an extern inline function always refers to the same object.

Note that the standard is explicitly talking about an "extern inline function" here.

> This works when everything is compile by one compiler but it does not work when m_a.cpp is compiled with clang and other files are compiled with g++:

Yes, the example you provided is the static local anonymous union case, which I agree needs to be fixed. Can you reupload a new patch which is limited to handling static local anonymous unions in inline functions, and avoid changes to non-ABI relevant names?

http://reviews.llvm.org/D6295