On 2020/01/30 08:37:44, hanwenn wrote: > There are two problems that are tangentially related, and I think the discussion > comes from us wanting to solve different problems. > > My problem is that it is very distracting to change the C++ code base if there > are compiler warnings about existing code. So, I want a code base that compiles > with zero warnings as a starting point. > > There is a different problem, which that there are some dubious casts in our > source code. I think that you, ie. Dan and David, want to address that problem. > > Here is a proposal that can make both of us happy: > > 1) fix all warning provisorially, by adding the casts that the compiler does > today explicitly. We mark them with "TODO: investigate cast". Result: the > compile becomes warning-free > > 2) go over all the "investigate cast" warnings, changing return types/signatures > where necessary. Result: our casts are now all kosher. > > I'm happy to contribute the work for 1). > > How does that sound?
I don't like this methodology, what's the difference over disabling -Wconversion? (which I don't think would be a good idea) I prefer what Dan does right now (btw thanks for working on this!): Pick a warning, investigate the issue and fix it correctly. Otherwise I see the risk that the TODOs will stick for longer than they should. So I think it's a good idea to keep the "distracting" warnings in place until they are correctly addressed. https://codereview.appspot.com/557190043/
