This patch rewrites checks in a few debug info tests to avoid using
'CHECK-NOT: {{DW_TAG|NULL}}'. It proposes --impicit-check-not=DW_TAG
instead, as it makes the checks clearer, and easier to analyze and update.
Details
Diff Detail
- Repository
- rG LLVM Github Monorepo
Event Timeline
llvm/test/DebugInfo/X86/align_c11.ll | ||
---|---|---|
27–33 | Perhaps this test could avoid testing all these other DW_TAGs by using --name or some other filtering on the dump, so it only dumps the relevant DIEs? | |
llvm/test/DebugInfo/X86/align_cpp11.ll | ||
81–84 | Any chance some filtering could be done to avoid checking a bunch of other tags in this test? |
Make some tests to filter entities by name to reduce the number of auxiliary checks.
llvm/test/DebugInfo/X86/align_cpp11.ll | ||
---|---|---|
81–84 | This test checks almost all entities from the dump, so filtering by name doesn't do much for it. Will update for comparing. |
Fair - I think the c11 test is better with the name query form. -your choice on the other one.
llvm/test/DebugInfo/X86/align_cpp11.ll | ||
---|---|---|
81–84 | Fair enough - yeah, maybe in future tests we can intentionally name the relevant entities to make them easy to match by name, but for this one I agree it's probably more name queries than is especially helpful. It is nice that this makes the test less dependent on the structure of the rest of the DWARF (eg: if we change the order that basic types get emitted for whatever reason - the name-query based version of the test is resilient to that in a way that the check-all-tags form is not) |
llvm/test/DebugInfo/X86/align_cpp11.ll with name filters looks a bit better to me, so I decided to keep it.
@dblaikie thank you for the review!
Perhaps this test could avoid testing all these other DW_TAGs by using --name or some other filtering on the dump, so it only dumps the relevant DIEs?