[flang-commits] [flang] [llvm] [Flang][OpenMP] Avoid aborting on nested derived types in DO CONCURRENT device conversion (PR #218963)

[via flang-commits]

https://github.com/MattPD commented:

Verified against `4395aa73662386f3ac0fda5534237618867c29fd`.

A small general suggestion: Would you be open to revising the PR message slightly so it focuses more on the patch's intent and rationale?

The PR message includes several review-time implementation details: the exact helper name (`fir::isRecordWithAllocatableMember`), the change that supplied the mangler guard, and that 26 tests ran (currently 25 runnable tests?). These details can be useful during review, but they may age quickly in the permanent message.

LLVM uses the PR body as the squash commit message. For more ideas on how to emphasize why and provide useful context, take a look at [LLVM's commit-message guidance](https://llvm.org/docs/DeveloperPolicy.html#commit-messages). It's always possible to mention the implementation details in the diff and review comments (if necessary for discussion). At the same time, future maintainers may get more value from understanding why flat mapping is sufficient for nested records without allocatable components, why array-valued record components need the same element-type traversal as mapper generation, and which mapping cases remain outside this change. Replacing the review-time details with that reasoning would make the history easier to understand without lengthening the description.


https://github.com/llvm/llvm-project/pull/218963