This test case captures the current state of support for printing branch targets.
Details
Diff Detail
- Repository
- rG LLVM Github Monorepo
Unit Tests
Time | Test | |
---|---|---|
190 ms | x64 debian > LLVM.Bindings/Go::go.test |
Event Timeline
llvm/test/tools/llvm-objdump/ELF/RISCV/branches.s | ||
---|---|---|
3 | Remove --match-full-lines In most test/tools/, the prevailing continuation-line format is to place | at the line end. # RUN: llvm-mc -filetype=obj -triple riscv32 -mattr=+c %s | # RUN: llvm-objdump -d -M no-aliases --no-show-raw-insn - | | |
5 | The address formatting is checked by other tests so this test doesn't need to check it. By removing addresses, when the number of instruction changes, the update (ongoing maintenance) will touch fewer lines. | |
7 | For similar reasons, remove leading addresses. | |
8 | I have a slight preference that instructions are indented by two: to be clearer that they belong to the function (nearest label). | |
77 | ||
79 | 60: c.nop should probably keep its address to be clear that its a jump target. |
Thank for such a rapid review @MaskRay! I've left a couple of queries in response, but otherwise I'll refresh the test case once I've had feedback on other patches in the series (to minimise repeated rebase effort).
llvm/test/tools/llvm-objdump/ELF/RISCV/branches.s | ||
---|---|---|
3 | The reasoning behind --match-full-lines was that it will catch changes to printing of branch targets (as in D116678), ensuring the test case is updated. | |
8 | Do you still feel it's worth doing, when taking into account the changes to this test in D116678? |
llvm/test/tools/llvm-objdump/ELF/RISCV/branches.s | ||
---|---|---|
3 | An alternative is to add {{$}} in a few places where it is significant. |
Remove --match-full-lines
In most test/tools/, the prevailing continuation-line format is to place | at the line end.