Details
- Reviewers
dblaikie aprantl dexonsmith - Commits
- rGeeb56abe6439: Update Clang for D20147 ("DebugInfo: New metadata representation for global…
rC281285: Update Clang for D20147 ("DebugInfo: New metadata representation for global…
rL281285: Update Clang for D20147 ("DebugInfo: New metadata representation for global…
Diff Detail
- Repository
- rL LLVM
Event Timeline
One thing that I forgot to do here was to add a test covering my changes to CGDebugInfo::EmitGlobalVariable. I'll do that momentarily.
| lib/CodeGen/CGDebugInfo.cpp | ||
|---|---|---|
| 3393 ↗ | (On Diff #57809) | Sorry about the confusion, I should have read the other patch first. This has morphed into the DIExpression. That's fine! |
| lib/CodeGen/CGDebugInfo.cpp | ||
|---|---|---|
| 3477 ↗ | (On Diff #57809) | It looks like we never really supported floating point constants in the DWARF output. I can only see support for integer constants: |
| lib/CodeGen/CGDebugInfo.cpp | ||
|---|---|---|
| 3477 ↗ | (On Diff #57809) | Looking at http://llvm-cs.pcc.me.uk/test/DebugInfo/X86/float_const.ll we at least do to some degree :-) |
| lib/CodeGen/CGDebugInfo.cpp | ||
|---|---|---|
| 3477 ↗ | (On Diff #57809) | That isn't a global variable test case though, is it? |
| lib/CodeGen/CGDebugInfo.cpp | ||
|---|---|---|
| 3477 ↗ | (On Diff #57809) | Ah that's right. We didn't support global float constants. |
LGTM with small changes.
| lib/CodeGen/CGDebugInfo.cpp | ||
|---|---|---|
| 3427 ↗ | (On Diff #57809) | I think it would be more readable to pass an empty DIExpression() here. |
| 3474 ↗ | (On Diff #57809) | Shouldn't the default constructor of DIExpression() wrap an MDNode nullptr? |
| lib/CodeGen/CGDebugInfo.h | ||
| 348 ↗ | (On Diff #57809) | While we're here: Emit *a* global variable's debug info? |