mizvekov added a comment. In D131858#3817750 <https://reviews.llvm.org/D131858#3817750>, @davrec wrote:
> Question was referring to @ChuanqiXu 's comment about this patch breaking > modules for function importing, see his repro (which I haven't looked at > closely). I was asking if this meant that classes were already broken in a > similar way, since you said this patch brings them into consistency. No, I haven't had time to investigate that test case, as right now I am advancing in other parts of the stack. But what I meant was that importing classes was already a case where the template parts could be accessed before the whole class was imported, and the existing code was already importing the template bits first to cope with that. The changes in this patch align the other cases to align with the need to import template parts first. > I think we misunderstand each other again, but are you saying the breakage > found by @ChuanqiXu does not need to be fixed because it is out of scope? > (If classes are already broken due to this issue, the argument might have > some validity.) No, fixing that breakage is in scope. What I mean is out of scope is a major refactoring that would change how objects are deserialized, so that they are not accessed while they are part way there. This, or a subset thereof, is what @ChuanqiXu was referring to, which would take about two months to implement. Repository: rG LLVM Github Monorepo CHANGES SINCE LAST ACTION https://reviews.llvm.org/D131858/new/ https://reviews.llvm.org/D131858 _______________________________________________ cfe-commits mailing list cfe-commits@lists.llvm.org https://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits
