On Thu, 26 Sep 2024 21:25:07 GMT, Phil Race <p...@openjdk.org> wrote:
>> There is some old awt/2d coding where warnings occur when running with ubsan >> enabled binaries. >> However at most of these locations the coding should work (at least on our >> supported platform set) so the warnings can be disabled at least for now. >> >> The change adds a macro ATTRIBUTE_NO_UBSAN similar to what we already use >> in Hotspot coding. > > This needs careful thinking about by various parties. > I am not sold on it, and the build team have indicated they do not support > --enable-ubsan and my experience with that and similar options is that they > are a nightmare to find a system on which they build and produce a working > binary. > > Until there's solid committed support by the leads of the build team I don't > think we should be considering these kinds of PRs. @prrace > the build team have indicated they do not support --enable-ubsan and my > experience with that and similar options is that they are a nightmare to find > a system on which they build and produce a working binary. The UBSan functionality is well supported, as such, in the build system. As you say, actually building and running a JDK with UBSan functionality is not trivial. This just means that we are unlikely to e.g. run it continuously on our CI in the foreseeable future, not that any changes towards improving support for detecting undefined behavior should be stopped. ------------- PR Comment: https://git.openjdk.org/jdk/pull/21184#issuecomment-2399519251
