On Wed, 4 Dec 2019 at 16:06, Thiago Macieira <thiago.macie...@intel.com> wrote: > > On Tuesday, 3 December 2019 17:36:37 PST Christian Gagneraud wrote: > > My point is that this sort of bugs are really hard to detect during > > code review. This is subtle & nasty 'feature' of QStringBuilder (a > > choice in its implementation). > > Understood, but it shows up quickly when running the code or, at worst, > Valgrind or ASan time.
Unfortunately, clang tooling is sporadically used here, but i'm pushing hard for that, dinosaurs move slowly. Since i started to use clang sanitiser, i found valgrind to be an historical technique. ASan, UBSan, TSan, ... have turned a massive amount of problems and the runtime impact is far less than Valgrind. As i was saying in the other email, we definitely don't live in the same world. I have to deal with "Fixing an historical bug will likely bring regression" sort of situation. That might sound crazy, but that is my everyday reality. Actually, Qt is full (has some) of that. I remember proposing a fix to QGraphicsView that was refused because the "bug" was considered a "feature", fixing the bug would break existing Qt client's code... Yeah, definitely Qt is full of that sort of things, you're not different in that respect. > It hasn't affected us in Qt itself because we use the Almost Never Auto coding > style. Yes, either Qt or Clang code guideline states that auto should only be used if the type is explicitly specified on the right hand side. The 'auto' feature was what that got me started with transitioning to C++11, I found it attractive and adopted it, a bit too much.... Lesson learned! Chris _______________________________________________ Interest mailing list Interest@qt-project.org https://lists.qt-project.org/listinfo/interest