This change is almost fully mechanical. The only interesting change is in generate_feature_test_macro_components.py to generate _LIBCPP_STD_VER >= instead. To avoid churn in the git-blame this commit should be added to the .git-blame-ignore-revs once committed.
Details
- Reviewers
ldionne Mordante var-const huixie90 - Group Reviewers
Restricted Project - Commits
- rG4f15267d3dd7: [libc++][NFC] Replace _LIBCPP_STD_VER > x with _LIBCPP_STD_VER >= x
Diff Detail
- Repository
- rG LLVM Github Monorepo
Event Timeline
LGTM, but I'll leave the approval to @ldionne (I remember this was discussed before and want to make sure this reflects the final decision; personally, I fully support this and think it's long overdue).
libcxx/utils/generate_feature_test_macro_components.py | ||
---|---|---|
286 | Out of curiosity, did we decide (or should we) to do the same for TEST_STD_VER? It seems slightly out of place seeing it use >. WDYT? |
libcxx/utils/generate_feature_test_macro_components.py | ||
---|---|---|
286 | I agree, however that touches even more files. Let's do it separately if at all. |
AFAIK we never made a decision :-(
I really would like to discuss this upfront instead of rushing it in. I expect merge conflicts in is some of my formatting patches. Not hard to fix, but it's quite annoying.
Mark and I talked about this offline. So my recollection is that we did have a discussion at some point on a review (or on Discord) where we had agreed to do this. I think I was initially against but had ended up agreeing since most folks wanted to do it. Clearly, this wasn't unanimous.
Mark and I agreed to set up an infrequent meeting with regular libc++ contributors to discuss these sorts of broad "policy" decisions in the future to make sure we had a clear agreement on those beforehand. We would take a few notes and publish them to Discourse or add them to a design document of sorts in the repository, making sure that those decisions are recorded somewhere.
Out of curiosity, did we decide (or should we) to do the same for TEST_STD_VER? It seems slightly out of place seeing it use >. WDYT?