sjoerdmeijer wrote: @nikic: I feel that all nuance has been lost here in this discussion, so I would like to bring some of that back: - First of all, let's recognise that Fortran and C/C++ are really different, and that they are different users of DA. At the time of enablement here, the last known bug that is being worked on involved type-punning of base pointers in a loop body, and/or some variants of this. You can't write that in Fortran. And in C/C++ you won't find this in normal code (it's also non-portable code). - In a bug triage process, before any action is taken, first the trigger conditions should be determined, and then the impact to determine the severity. We don't have any of this information. I thus feel we have different standards for DA compared to other components that have problems (that may or may not have an impact). We should also recognise that new information appeared after this patch was merged, and very recently. I - I haven't looked at all new bugs, but one of them is definitely another very weird corner case that I doubt can be triggered from Fortran. And whether the exact process has been followed for enablement I don't know, but now the fact is that this has been running for 3 months and no bugs have been raised against Flang or interchange.
If you feel this should be reverted based on the grounds of process or inclusion, okay, then it is what it is, fair enough. Going forward though I would encourage a more constructive approach to deal with bugs before any conclusions are drawn. https://github.com/llvm/llvm-project/pull/140182 _______________________________________________ cfe-commits mailing list cfe-commits@lists.llvm.org https://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits
