baloghadamsoftware added a comment. In https://reviews.llvm.org/D54466#1297887, @NoQ wrote:
> > I marked this patch as WIP because I could not create a test-case for it. > > However in real projects this patch seems to reduce false positives > > significantly. > > False positives are hard to reduce via delta-debugging because they can be > accidentally reduced into true positives, because the distinction between > true positives and false positives cannot be tested automatically. But once > you fix the false positive, `creduce` can be used to reduce them by writing > down the condition as "the positive is there on the original clang but not on > the modified clang". Strongly recommended :) `creduce` did not work (continous `clang_delta` crash) but somehow I managed to create tests. >> Hmmm, shouldn't we add this to `MemRegion`'s interface instead? This: > I wouldn't insist, but this does indeed sound useful. I suggest > `MemRegion::getMostDerivedObjectRegion()` or something like that. vs > Also, ugh, that nomenclature: the base region of `CXXBaseObjectRegion` in > fact represents the //derived// object. So if `CXXBaseObjectRegion` is the derived object, then `MemRegion::getMostDerivedObjectRegion()` gets the least derived object region now? https://reviews.llvm.org/D54466 _______________________________________________ cfe-commits mailing list cfe-commits@lists.llvm.org http://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits
