This PR consists of 2 parts. The first change is to use addr_of_mut! when getting pointers to the EMPTY_CONSOLE, which fixes some warnings about mutable references to static mut objects on recent nightlies.
The second change is one I'm making based more on "vibes" than anything, but I've long felt suspicious about the soundness of our Console API when it comes to taking references and pointers to the internal PrintConsole. For example, we have a number of console APIs that take &self even though there are underlying mutations in the PrintConsole struct every time a character is written to the screen.
I've never actually run into any problems or strange behavior when dealing with the console, but I suspect that the "proper" way to handle things is to wrap the PrintConsole in an UnsafeCell so that the compiler won't make assumptions about having unique access to the console state unless we specifically tell it to do so, and so I've made that change in this PR too.
It's possible that this change might be overly paranoid, but I'm not sure there's any harm in being paranoid in this particular way either. Either way, feel free to let me know what you guys think about the issue too.
This PR consists of 2 parts. The first change is to use
addr_of_mut!when getting pointers to theEMPTY_CONSOLE, which fixes some warnings about mutable references tostatic mutobjects on recent nightlies.The second change is one I'm making based more on "vibes" than anything, but I've long felt suspicious about the soundness of our
ConsoleAPI when it comes to taking references and pointers to the internalPrintConsole. For example, we have a number of console APIs that take&selfeven though there are underlying mutations in thePrintConsolestruct every time a character is written to the screen.I've never actually run into any problems or strange behavior when dealing with the console, but I suspect that the "proper" way to handle things is to wrap the
PrintConsolein anUnsafeCellso that the compiler won't make assumptions about having unique access to the console state unless we specifically tell it to do so, and so I've made that change in this PR too.It's possible that this change might be overly paranoid, but I'm not sure there's any harm in being paranoid in this particular way either. Either way, feel free to let me know what you guys think about the issue too.