[clang] [WIP][analyzer] Refactor `ExplodedGraph::trim()` (PR #139939)

Balazs Benics via cfe-commits cfe-commits at lists.llvm.org
Thu May 15 06:46:21 PDT 2025


=?utf-8?q?Donát?= Nagy <donat.nagy at ericsson.com>,
=?utf-8?q?Donát?= Nagy <donat.nagy at ericsson.com>,
=?utf-8?q?Donát?= Nagy <donat.nagy at ericsson.com>,
=?utf-8?q?Donát?= Nagy <donat.nagy at ericsson.com>,
=?utf-8?q?Donát?= Nagy <donat.nagy at ericsson.com>,
=?utf-8?q?Donát?= Nagy <donat.nagy at ericsson.com>,
=?utf-8?q?Donát?= Nagy <donat.nagy at ericsson.com>,
=?utf-8?q?Donát?= Nagy <donat.nagy at ericsson.com>,
=?utf-8?q?Donát?= Nagy <donat.nagy at ericsson.com>,
=?utf-8?q?Donát?= Nagy <donat.nagy at ericsson.com>
Message-ID:
In-Reply-To: <llvm.org/llvm/llvm-project/pull/139939 at github.com>


steakhal wrote:

> I pushed a heap of small commits which do various simple cleanup in this neighborhood.
> 
> > If we could harden the bug EQclass part for deterministically selecting the sequence of the bug reports we would try, that would make the outcomes stable no matter how the egraph looks, right?
> > Anyway, we should probably focus first hardening stability of the reports and then on disruptive changes to mitigate the pain upfront. Because if we would land this, and then more similar changes, then we would face this instability again and again with a similar magnitude.
> 
> I agree, this seems to be a good approach: first land one commit that enforces deterministic ordering of the bug reports (which is non-NFC and may disrupt the users _once_) and then it will be possible to merge this commit and other similar changes without further disruptions.
> 
> I guess, then this commit is put on hold until you or me or somebody else can harden the ordering of bug reports within an EQclass... I'm not sure that I can do so immediately (my backlog of plans is getting larger each day...) but there is a chance...

I think we are aligned. yes.

https://github.com/llvm/llvm-project/pull/139939


More information about the cfe-commits mailing list