4 ms·
I daresay that some codebases might rely on overflowing behaviour even if they don't know it. I've worked with a codebase where assumptions about unused RAM wer
by pyre 11y ago
I daresay that some codebases might rely on overflowing behaviour even if they don't know it. I've worked with a codebase where assumptions about unused RAM were that it was always zeroed out. Moving to a different platform (Linux) changed those assumptions, and there was lots of weird behaviour in the codebase because of that.
I'll bet there are lots of legacy codebases that would actively resist using such a compiler technology for C/C++ because of unknown behaviours happening in their codebase (and because there is no test suite, or just poor test coverage).
- dmix 11y agoLegacy codebases like that tend to run on outdated operating systems, so addressing issues with buffer overflows in the software would probably do little to improve the security of the systems as a whole.
- dllthomas 11y agoCertainly, though usually that winds up perceived as brittleness moving between compiler versions in general. Ideally, the tech we are discussing could help make it easier to understand those unknown behaviors. But even if not, and even if this accounts for 60% of existing C/C++ codebases, improving the remaining 40% is a huge win that would take tremendously more time to reimplement in Rust. This is not to say that I am not also very excited about Rust.