RFR: 8365604: Null pointer dereference in src/hotspot/share/adlc/output_h.cpp ArchDesc::declareClasses() [v2]

[Artem Semenov]

On Thu, 21 Aug 2025 14:32:43 GMT, Andrew Dinn <adinn at openjdk.org> wrote:

> Well, this leads right to the root of the problem I have with this report. As you say, pos_idx does indeed come out of a marker object. It took me about a minute to identify that this marker object is created in the function that sits right above the one your code assistant flagged as problematic -- even though I am not at all familiar with this code. It looks clear to me that, given the right call sequence for calls that create a marker and then consume it here, the check on pos_idx will ensure that we don't drop off the end of the list with a null pointer. So, it looks very liek this code has been designed so that the presence of a marker with a suitable pos_idx is intended to ensure this loop terminates before that happens. I am sure someone in this project knows whether that is the case but it is not you or your coding assistant.
> 
> I'm not suggesting that that calling sequence is actually right and that the check for pos_idx will definitely avoid dropping off the end. Indeed, I would welcome a bug report that proved it to be wrong. However, what is clear that both you and your coding assistant have failed to appreciate how some relatively obvious parts of this design actually operate. That renders your (or your tool's) analysis a shallow and unhelpful distraction; using it as an excuse to raise a purported 'issue' in the absence of any evidence of an actual issue is very much a waste of time for this project's reviewers.
> 
> Your error is compounded by the fact that you (or more likely your coding assistant) are suggesting changes which, because they are not founded in a correct understanding of the design, could potentially lead to worse outcomes than the speculative nullptr dereference they are intended to remedy -- as I explained when discussing your change to the control flow logic in the ALDC code. So, not only is this report unhelpful it is potentially harmful.
> 
> Ultimately the takeaway here is that the OpenJDK bug system is not here to report, review and add patches to remedy issues that you or your code assistant tool invents on the basis of misinformed assumptions. It is here to report, review and add patches to remedy issues that can be shown to actually affect the correct operation of the JVM and JDK,either by a reproducible test or by well-reasoned argument. So, please do not continue to spam the project with bug reports like this simply because a potentially bogus patch will improve your experience with what is clearly a decidedly fallible tool.

I’m sorry to have taken up your time.

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

PR Review Comment: https://git.openjdk.org/jdk/pull/26798#discussion_r2293532235