RFR: 8333277: ubsan: mlib_ImageScanPoly.c:292:43: runtime error: division by zero

[Matthias Baesken]

On Wed, 12 Jun 2024 16:11:01 GMT, Phil Race <prr at openjdk.org> wrote:

> We'll never be ubsan clean and I'm not sure it is advisable in all cases.

I know about a number of cases in HS coding where we do things that ubsan warns about (e.g. triggering various signals/crashes) but those things are still valid. But so far those cases are few and can be handled (e.g. switching off an ubsan check subclass or switching off ubsan checking for the function/method with the specific code).

Having the ubsan support in configure already for some time, it is at least desired to have a working OpenJDK build when '--enable-ubsan' is set;  we are not there yet but very close, a few places in HS remain to be adjusted.
For the whole codebase it is currently not very likely that we get 'ubsan clean'  (whatever 'ubsan clean' may mean).
For e.g.  ':tier1' in HS / JDK it might be achievable with medium effort (not so many issues remaining).

-------------

PR Comment: https://git.openjdk.org/jdk/pull/19674#issuecomment-2165803918