On 05/25/12 21:36, Jonathan M Davis wrote: > On Friday, May 25, 2012 13:52:05 simendsjo wrote: >> Couldn't dmd automatically add an assert(c) in debug mode before accessing >> members? > > That's been argued for ages. Walter is against it, because the OS already > takes care of it with segfaults. you can either run the program again (which > isn't necessarily true, but I believe with the programs that he seems to run > normally, it is, so he's got somewhat of a skewed perspective on that), or if > you have core dumps enabled, you can just examine the core dump to see what > happened. Also, adding all of those assertions would also negatively impact > non-release mode - and possibly by quite a bit, since it would be happening > all over the place, so there's that to consider. > > Regardless, it's a rather devisive subject. I think that _most_ people would > like to have such assertions added in non-release mode, but since Walter's > against it, it's never happened.
It's bad enough that such check is forced by the compiler in non-release mode when calling methods. That means you get to choose between not supporting non-release builds at all, or having an awkward, unnatural API, for no other reason than this compiler limitation. Checking for non-null refs/pointers before accessing members would actually be less problematic, as dereferencing null is *always* a bug, unlike the method case, where it can be perfectly fine. Tuning this behavior on a per-type basis via attributes would be best, but that's not likely to happen soon, i guess. artur
