[llvm-bugs] [Bug 38607] New: Using a private global as a global_ctor key fails when emitting a COFF object
via llvm-bugs
llvm-bugs at lists.llvm.org
Thu Aug 16 13:30:42 PDT 2018
https://bugs.llvm.org/show_bug.cgi?id=38607
Bug ID: 38607
Summary: Using a private global as a global_ctor key fails when
emitting a COFF object
Product: libraries
Version: trunk
Hardware: PC
OS: Windows NT
Status: NEW
Severity: enhancement
Priority: P
Component: Common Code Generator Code
Assignee: unassignedbugs at nondot.org
Reporter: rnk at google.com
CC: david.majnemer at gmail.com, llvm-bugs at lists.llvm.org
Consider this LLVM IR:
@x = private global i32 0
@llvm.global_ctors = appending global [1 x { i32, void ()*, i8* }]
[{ i32, void ()*, i8* }
{ i32 65535,
void ()* @register_x,
i8* bitcast (i32* @x to i8*) }]
define internal void @register_x() {
ret void
}
We can generate asm for it, but we cannot assemble it to an object. We emit:
$ llc t.ll -o t.o -filetype=obj
LLVM ERROR: Missing associated COMDAT section for section .CRT$XCU
Mechanically, what happens is we try to look up a private label (.Lx) in the
COFF symbol table that we're making, we use GetOrCreate on it, but we don't
have a section for it, so we issue this diagnostic.
However, the input LLVM IR makes some sense. We should run the initializer if
and only if @x is included in the final link, and it's not comdat, so it will
be included, and we can just ignore it and always put @register_x in the main
.CRT$XCU section.
The question is, should we accept the assembly that we generate today?
.section .rdata,"dr"
.p2align 2 # @x
.Lx:
.long 0 # 0x0
.section .CRT$XCU,"dr",associative,.Lx
.p2align 3
.quad register_x
Does that assembly make sense, i.e. make this .CRT$XCU section associative with
the non-comdat .rdata section?
--
You are receiving this mail because:
You are on the CC list for the bug.
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.llvm.org/pipermail/llvm-bugs/attachments/20180816/8b2cb0a4/attachment.html>
More information about the llvm-bugs
mailing list