[llvm] [DebugInfo] DWARFFormValue use formatv instead of format (PR #180498)

[Konrad Kleine via llvm-commits]

================
@@ -552,19 +557,20 @@ void DWARFFormValue::dump(raw_ostream &OS, DIDumpOptions DumpOpts) const {
     break;
 
   case DW_FORM_rnglistx:
-    OS << format("indexed (0x%x) rangelist = ", (uint32_t)UValue);
+    OS << formatv("indexed ({0:x+}) rangelist = ", (uint32_t)UValue);
     break;
 
   case DW_FORM_loclistx:
-    OS << format("indexed (0x%x) loclist = ", (uint32_t)UValue);
+    OS << formatv("indexed ({0:x+}) loclist = ", (uint32_t)UValue);
     break;
 
   case DW_FORM_sec_offset:
-    AddrOS << format("0x%0*" PRIx64, OffsetDumpWidth, UValue);
+    AddrOS << formatv(
+        "0x{0:x-}", fmt_align(UValue, AlignStyle::Right, OffsetDumpWidth, '0'));
----------------
kwk wrote:

`0x{0:x-}` here ensures that the alignment happens and the `0x` is prefix. When I use `{0:x+}` for example to add the `0x` prefix automatically, the alignment considers it to be part of the string to prefix with `'0'`s. In other words you get examples like these `0000000xff` which is most certainly not what one wants.

The other places where I've just used `{0:x-}` the original format didn't include a `0x` prefix either. For example:

```diff
-        AddrOS << format("indexed (%8.8x) address = ", (uint32_t)UValue);
+        AddrOS << formatv("indexed ({0:x-8}) address = ", (uint32_t)UValue);
```

For example

```c++
format("%2.2", 255) # yields: ff
formatv("{0:x-2}", 255) # yields: ff
```

https://github.com/llvm/llvm-project/pull/180498