http://llvm.org/bugs/show_bug.cgi?id=21111
David Majnemer <[email protected]> changed: What |Removed |Added ---------------------------------------------------------------------------- Status|NEW |RESOLVED Resolution|--- |FIXED Assignee|unassignedclangbugs@nondot. |[email protected] |org | --- Comment #2 from David Majnemer <[email protected]> --- There are many things we could do to make this code compile and execute correctly. Off of the top of my head: - We could export the tls_index that the variable lives in and utilize that when we reference the variable. - We could create a thunk for the variable in all translation units which define the variable. The underlying variable would be thread local but not exported; instead, we would export the thunk. All access to the variable would go through the wrapper. This is strikingly similar to the Itanium thread-wrapper ABI. Seeing as how MSVC will not accept this code, I see no reason to go through heroics to make this work. Fixed in r219049. -- You are receiving this mail because: You are on the CC list for the bug.
_______________________________________________ LLVMbugs mailing list [email protected] http://lists.cs.uiuc.edu/mailman/listinfo/llvmbugs
