Re:Implications of Rewriting a Browser Component in Rust
by maelkum on Friday March 01, 2019 @06:22AM
(#19279576)
I have a codebase that amounts to several kLOC of C++, available here: https://git.sr.ht/~maelkum/viuavm While I openly admit that parts of it are shoddily written, I also try to always use the "reliable subset" of C++.
From my experience it is not hard to make it a habit (e.g. using `::at()` and declaring variables as `auto const x = ...;` is muscle memory at this point) but the code becomes very verbose and looks like programmed defensively to a sometimes ridiculous extent.
So I understand your skepticism. To address the point of relying on developers being "ideal": just turn the compiler warning flags up to eleven, make all warnings errors, and run your tests two times - one time with sanitisers, and the other under Valgrind. This won't catch all errors, sure, but will still make you more confident in your code's reliability.
If so, how much do they rely on the developers being ideal programmers, who know intimately the intricacies of C++ and how to avoid ending up in UB-land?