 Have you ever heard about LibreOffice breaking your toolchain for a long time? So I forgot to put the proper date on the slide. So the first thing we break is Android. Have you heard from Thor that Android is the new Windows? So the first problem is that GDP has exactly the same problems as Android today but used to be true on normal PC as well. Like if you would like to put a breakpoint in a function that's in a sharp library that's not yet loaded, then you can do that. So you have to, these are just quotes from Redmi phones. So you need to carry forward the guess where to put the breakpoint and once you guess where the bug is, you can prove the breakpoint there to fix the bug. Maybe there is some logical loop there. And is this actually still true or solved today? It's not yet solved, yeah. So this is still true. The other one is when the Android link breaks and the solution for this is that some of just the code output changed enough so that it no longer happens. So the Tinder was just tripped itself with waiting for enough time. So it's some internal error in the code linker. But of course not only Android is broken. So this must be seen, the Microsoft compiler. So this is the output from the compiler for the fourth version of Kandy's reward right of you at our map-free switch cases. I guess initially it wasn't switch case but it was instead some huge list of conditions. And if you have enough nested conditions then GCC and Clang is happy but not the Microsoft compiler. You don't have to use too much nesting. If you use a code generator then it's easy to do nesting with something like 200 levels and then the Microsoft compiler will already break. So let's break Clang. How about this workaround for iOS where in the middle of the write or layout rendering if you are doing an optimized build then we have to call this dummy function where we query the process ID of the current application. And if we do nesting with it then our actor layout gets fixed. Sadly Apple's Clang version is not open source as far as I know. So this is something we can just work around. The good thing is that for GCC we can report bugs. So it's really interesting if you start to search for libraries class names in the GCC bug track code. There are many hits there. So one example is we have another reference and you would like to just cast it to the real implementation to do something there. And one GCC bug was that if you just did the downcast to the basecast directly then it returned an R. But if you did it in two stops then it worked. So this is something that doesn't happen with today's GCC so this got fixed. And also this is just one more example when we had some problems and GCC guys managed to fix it or at least they took part in the process. At least today's GCC doesn't have the problem. I guess that's it in case you have any questions. No, then I am safe. Thanks for listening.