On Tue, 13 Dec 2022 16:55:09 GMT, Robbin Ehn <r...@openjdk.org> wrote:
> > I guess the advantage to putting this in the build machinery (as opposed to > > using `--with-extra-cflags=-fsanitize=undefined > > --with-extra-ldflags=-fsanitize=undefined`) is that we can turn some of > > these onn by default once we've fixed each category of UB. Is that right? > > It will take a while, look a bit on align issue, we have so much code which > go from pointer to large -> small -> large, e.g. > > ``` > address addr = data() + offset; > return (ImmutableOopMap*) addr; > ``` > > In this case data() needs to return something with the same alignment as a > ptr and offset must be in even in ptr steps. Yeah, some of the cases we may just have to suppress if they are not feasible to fix without lots of effort. My intention is to fix all the low hanging fruit, and then suppress the remaining cases. For the suppressed cases, I'll file P4 bugs for each one. Then if somebody feels like fixing them, they can do it. ------------- PR: https://git.openjdk.org/jdk/pull/11604
