| // LLDB 1800+ tests were not tested in CI, broke, and now are disabled | |
| //@ ignore-lldb | |
| //@ compile-flags:-g | |
| //@ disable-gdb-pretty-printers | |
| //@ ignore-backends: gcc | |
| // === GDB TESTS =================================================================================== | |
| //@ gdb-command:set print union on | |
| //@ gdb-command:run | |
| //@ gdb-command:print case1 | |
| //@ gdb-check:$1 = struct_style_enum::Regular::Case1{a: 0, b: 31868, c: 31868, d: 31868, e: 31868} | |
| //@ gdb-command:print case2 | |
| //@ gdb-check:$2 = struct_style_enum::Regular::Case2{a: 0, b: 286331153, c: 286331153} | |
| //@ gdb-command:print case3 | |
| //@ gdb-check:$3 = struct_style_enum::Regular::Case3{a: 0, b: 6438275382588823897} | |
| //@ gdb-command:print univariant | |
| //@ gdb-check:$4 = struct_style_enum::Univariant::TheOnlyCase{a: -1} | |
| // === LLDB TESTS ================================================================================== | |
| //@ lldb-command:run | |
| //@ lldb-command:v case1 | |
| //@ lldb-check:(struct_style_enum::Regular) case1 = Case1{a:0, b:31868, c:31868, d:31868, e:31868} { a = 0 b = 31868 c = 31868 d = 31868 e = 31868 } | |
| //@ lldb-command:v case2 | |
| //@ lldb-check:(struct_style_enum::Regular) case2 = Case2{a:0, b:286331153, c:286331153} { a = 0 b = 286331153 c = 286331153 } | |
| //@ lldb-command:v case3 | |
| //@ lldb-check:(struct_style_enum::Regular) case3 = Case3{a:0, b:6438275382588823897} { a = 0 b = 6438275382588823897 } | |
| //@ lldb-command:v univariant | |
| //@ lldb-check:(struct_style_enum::Univariant) univariant = TheOnlyCase{a:-1} { a = -1 } | |
| #![allow(unused_variables)] | |
| use self::Regular::{Case1, Case2, Case3}; | |
| use self::Univariant::TheOnlyCase; | |
| // The first element is to ensure proper alignment, irrespective of the machines word size. Since | |
| // the size of the discriminant value is machine dependent, this has be taken into account when | |
| // datatype layout should be predictable as in this case. | |
| enum Regular { | |
| Case1 { a: u64, b: u16, c: u16, d: u16, e: u16}, | |
| Case2 { a: u64, b: u32, c: u32}, | |
| Case3 { a: u64, b: u64 } | |
| } | |
| enum Univariant { | |
| TheOnlyCase { a: i64 } | |
| } | |
| fn main() { | |
| // In order to avoid endianness trouble all of the following test values consist of a single | |
| // repeated byte. This way each interpretation of the union should look the same, no matter if | |
| // this is a big or little endian machine. | |
| // 0b0111110001111100011111000111110001111100011111000111110001111100 = 8970181431921507452 | |
| // 0b01111100011111000111110001111100 = 2088533116 | |
| // 0b0111110001111100 = 31868 | |
| // 0b01111100 = 124 | |
| let case1 = Case1 { a: 0, b: 31868, c: 31868, d: 31868, e: 31868 }; | |
| // 0b0001000100010001000100010001000100010001000100010001000100010001 = 1229782938247303441 | |
| // 0b00010001000100010001000100010001 = 286331153 | |
| // 0b0001000100010001 = 4369 | |
| // 0b00010001 = 17 | |
| let case2 = Case2 { a: 0, b: 286331153, c: 286331153 }; | |
| // 0b0101100101011001010110010101100101011001010110010101100101011001 = 6438275382588823897 | |
| // 0b01011001010110010101100101011001 = 1499027801 | |
| // 0b0101100101011001 = 22873 | |
| // 0b01011001 = 89 | |
| let case3 = Case3 { a: 0, b: 6438275382588823897 }; | |
| let univariant = TheOnlyCase { a: -1 }; | |
| zzz(); // #break | |
| } | |
| fn zzz() {()} |