[MC][WebAssembly] Error on data symbols in the text section.

[sbc100]

Summary

Previously we had an assert but this can actually occur in valid user
code so we need to handle this in release builds too.

[dschuff]

Is there any easy way to print the name of the symbol? One thing I discovered when working with a user with a big codebase is that these backend error messages we have aren't very useful, even for us who know exactly what they mean.

[sbc100]

This does include the name of symbol in the error. I admit there are other place in the binary reader where we don't do that.

[dschuff]

Oh, right. Guess I should actually look at the code first...

llvm/lib/MC/WasmObjectWriter.cpp

if (WS.isFunction()) {
         report_fatal_error("data symbols must have a size set with .size: " +
                            WS.getName());
 
       int64_t Size = 0;
       if (!WS.getSize()->evaluateAsAbsolute(Size, Layout))
         report_fatal_error(".size expression must be evaluatable");
 
       auto &DataSection = static_cast<MCSectionWasm &>(WS.getSection());
-      assert(DataSection.isWasmData());
+      if (!DataSection.isWasmData())
+        report_fatal_error("data symbols must live in a data section: " +
+                           WS.getName());
 
       // For each data symbol, export it in the symtab as a reference to the
       // corresponding Wasm data segment.
       wasm::WasmDataReference Ref = wasm::WasmDataReference{
           DataSection.getSegmentIndex(),
           static_cast<uint32_t>(Layout.getSymbolOffset(WS)),
           static_cast<uint32_t>(Size)};
       DataLocations[&WS] = Ref;

llvm/test/MC/WebAssembly/data-symbol-in-text-section.ll

+; RUN: not llc -filetype=obj %s -o /dev/null 2>&1 | FileCheck %s
+; CHECK: data symbols must live in a data section: data_symbol
+
+target datalayout = "e-m:e-p:32:32-i64:64-n32:64-S128"
+target triple = "wasm32-unknown-unknown"
+
+@data_symbol = constant [1024 x i32] zeroinitializer, section ".text", align 16
+
+define hidden i32 @main() local_unnamed_addr #0 {
+entry:
+  %0 = load i32, i32* getelementptr inbounds ([1024 x i32], [1024 x i32]* @data_symbol, i32 0, i32 10)
+  ret i32 %0
+}