Inconsistent DWARF operation evaluation error message across LLVM versions when using -Oz #111041
Description
Inconsistent debug information error messages have been observed when evaluating certain DWARF operations. Specifically, when attempting to print the value of p_31 at line 17 in a reduced C code sample, different error messages are produced depending on the LLVM version used. Godblot link
In LLVM 19.1.0, a more detailed error message is generated: Could not evaluate DW_OP_entry_value: no matching call site param found
In contrast, older LLVM versions (18.1.8, 17.0.6, and 16.0.3) show a shorter error message: Could not evaluate DW_OP_entry_value
These discrepancies can be replicated by following certain steps when compiling and debugging the reduced C code sample. In LLVM19.1.0:
clang -g -Oz -o d1_Oz.out d1.c
(lldb) file d1_Oz.out
(lldb) b 17
(lldb) r
[...]
* thread #1, name = 'd1_Oz.out', stop reason = breakpoint 1.1
frame #0: 0x0000555555555162 d1_Oz.out`func_28(p_31=<unavailable>) at d1.c:17:1
14 }
15 int32_t *func_28(uint32_t p_31) {
16 *g_105 = p_31 * g_9;
-> 17 }
18 int16_t func_33() {
19 }
20 void main() {
(lldb) fr v
(uint32_t) p_31 = <could not evaluate DW_OP_entry_value: no matching call site param found>
(lldb) p p_31
error: Couldn't materialize: couldn't get the value of variable p_31: could not evaluate DW_OP_entry_value: no matching call site param found
error: errored out in DoExecute, couldn't PrepareToExecuteJITExpression
llvm-dwarfdum --debug-info d1_Oz.out
0x000000dd: DW_TAG_formal_parameter
DW_AT_location (indexed (0x1) loclist = 0x00000022:
[0x0000000000001152, 0x0000000000001159): DW_OP_reg5 RDI
[0x0000000000001159, 0x0000000000001163): DW_OP_entry_value(DW_OP_reg5 RDI), DW_OP_stack_value)
DW_AT_name ("p_31")
DW_AT_decl_file ("d1.c")
DW_AT_decl_line (15)
DW_AT_type (0x0000013c "uint32_t")
cat d1.c
#include "stdint.h"
int32_t g_9, g_106, g_38;
int32_t *g_105 = &g_106;
int8_t func_23(uint16_t);
int32_t *func_28();
int16_t func_33();
int16_t func_1() {
uint32_t l_12;
int32_t *l_218 = &g_38;
*l_218 = func_23(l_12);
}
int8_t func_23(uint16_t) {
func_28(func_33);
}
int32_t *func_28(uint32_t p_31) {
*g_105 = p_31 * g_9;
}
int16_t func_33() {
}
void main() {
func_1();
}