[DebugInfo] Generate fixups as emitting DWARF .debug_frame/.eh_frame.

[HsiangKai]

Summary

It is necessary to generate fixups in .debug_frame or .eh_frame as
relaxation is enabled due to the address delta may be changed after
linking.

There is an opcode with 6-bits data in debug frame encoding. So, we
also need 6-bits fixup types.

[jrtc27]

Perhaps we should keep getKindForSize as being in bytes, and add a new getKindForSizeInBits? This would reduce the diff, but more importantly ensures that downstream forks don't silently break until runtime due to the changed semantics of getKindForSize despite having the same type as before.

[jrtc27]

include/llvm/MC/MCFixup.h

162 ↗

(On Diff #187179)

Keep this on one line like the surroundings?

176 ↗

(On Diff #187179)

Ditto.

[HsiangKai]

include/llvm/MC/MCFixup.h

162 ↗

(On Diff #187179)

Formatted by clang-format. However, I agree with you. It should be kept the same format as previous lines.

[HsiangKai]

Ping. The patch is similar to D46850.

[HsiangKai]

Support .debug_frame/.eh_frame parsing for llvm-dwarfdump.

[asb]

This seems sensible to me from a RISC-V perspective, though I'd appreciate it if one of the debug info people could feedback on the DWARF related changes. Might you have a chance to look at this @aprantl @probinson @JDevlieghere?

[probinson]

The dwarfdump changes look okay, but the new tests don't exercise those changes. The simplest thing is probably to RUN llvm-dwarfdump in relax-debug-frame.ll and verify the output describes the frame as expected.

lib/MC/MCDwarf.cpp

1907 ↗

(On Diff #202409)

bool WithFixups = (Offset != nullptr && Size != nullptr);
or even just (Offset && Size)

test/DebugInfo/RISCV/relax-debug-frame.ll

8 ↗

(On Diff #202409)

Do you have a way to prove these relocations are in the right section? Right now the test verifies that they come after the start of the .rela.*_frame section, but it does not verify they come before the end of the section.

[HsiangKai]

Update test cases.

[HsiangKai]

Use llvm-dwarfdump to ensure DWARF parsing is correct under relaxation.

test/DebugInfo/RISCV/relax-debug-frame.ll

8 ↗

(On Diff #202409)

I check the end of section '}' before every pair of checking strings. It could ensure these strings within one section.

[HsiangKai]

Ping.

[asb]

LGTM (two minor comments in-line). Pleas go ahead and commit - thanks!

include/llvm/MC/MCFixup.h

147 ↗

(On Diff #206128)

isPCRel -> IsPCRel

153 ↗

(On Diff #206128)

Can change if(foo)+llvm_unreachable to an assert

[asb]

Looking at this in more detail - shouldn't we be making use of the R_RISCV_SET* relocations?