[llvm] [DebugInfo] Don't set prologue_end behind line-zero call insts (PR #156850)

[Stephen Tozer via llvm-commits]

================
@@ -0,0 +1,129 @@
+# RUN: llc %s -start-after=livedebugvalues -o - -filetype=obj | llvm-dwarfdump - --debug-line | FileCheck %s --implicit-check-not=prologue_end
+#
+## Original code, compiled clang -O2 -g -c
+## 
+## void ext();
+## int main(int argc, char **argv) {
+##   if (argc == 1)
+##     ext();
+##   else
+##     ext();
+##    return 0;
+## }
+## 
+## In the code sequence above, the call to ext is given line zero during
+## optimisation, because the code is duplicated down all function paths thus
+## gets merged. We get something like this as the output:
+##
+##   0:   50                      push   %rax
+##   1:   31 c0                   xor    %eax,%eax
+##   3:   e8 00 00 00 00          call   8 <main+0x8>
+##                        4: R_X86_64_PLT32       ext-0x4
+##   8:   31 c0                   xor    %eax,%eax
+##   a:   59                      pop    %rcx
+##   b:   c3                      ret
+##
+## Where prologue_end is placed on address 8, the clearing of the return
+## register, because it's the first "real" instruction that isn't line zero.
+## This then causes debuggers to skip over the call instruction when entering
+## the function, which is catastrophic.
+##
+## Instead: we shouldn't put a prologue_end on this function at all. It's too
+## deformed from the original code to truly have a position (with a line number)
+## that is both true, and after frame setup. This gives comsumers the
----------------
SLTozer wrote:

```suggestion
## that is both true, and after frame setup. This gives consumers the
```

https://github.com/llvm/llvm-project/pull/156850