On Saturday 24 November 2007, Carl-Daniel Hailfinger wrote:
I produce two classes of patches:
- Patches where I'm not sure they will work [...]
- Patches which are IMO obviously correct. They can involve pretty large
code changes or rewriting fragile (and long-time unchanged) code everybody uses. Obviously correct patches are trivial.
That is a courageous assumption. Code that looks correct is trivial? I miss the third class of patches like comment reformatting, re-indent, naming constant values and the like, changes every one of us would consider trivial.
I have no problem committing a rewrite of the CAR setup assembly code. It is obviously correct for me even without having ever looked at any of the relevant data sheets or having tested it on real hardware. Do you really want me to commit that? The patch is ready.
If that code does work in all cases I might consider you a genius, depending on the new code. If it breaks under some circumstances and you don't care, then this would indeed be the first step to have your commit rights removed, IMHO. If you're not feeling lucky, ask for a second opinion. It's that simple.