From: Jeff H. <jeffh@ActiveState.com> - 2003-08-14 04:03:58
|
John Ousterhout wrote: > 1. We care about having warning-free compiles. In this case, turning > warnings into errors is helpful, since it ensures that they get fixed > immediately. > > 2. We don't care about having warning-free compiles. In this case, > there's no need to turn them into errors. > > I'm having trouble coming up with a scenario where you'd want approach > #1 but wouldn't want warnings turned into errors. This feels like a > slippery slope where you'll end up with zillions of warnings. There is no intention on being more lenient on warnings. I think the reason that this was done on Windows was to ensure cleaner code, but also because it was possible to enforce this on a single strict API system that had a single target - x86 Win32 builds. The problem is that this no longer really applies, and you start to have the same issues that the *nix code endures from different variations. The win/ stuff can now target Win64 and Win/CE (both have multiple cpu targets), which can have niggling API differences. -- Jeff Hobbs The Tcl Guy Senior Developer http://www.ActiveState.com/ Tcl Support and Productivity Solutions |