Re: [PATCH v3 bpf-next] bpf: improve the general precision of tnum_mul

[Nandakumar Edamana <nandakumar@xxxxxxxxxxxxxxxx>]

I've checked the three new test programs on an old kernel that has the 
current tnum_mul and a newly built one that has the proposed tnum_mul. 
verifier_mul/mul_precise results in a false positive with the current 
tnum_mul, while the program passes with the new build. While it is 
unfair to compare two kernel versions, logs indicated this is indeed a 
multiplication-related issue.

Honestly, with some effort, creating examples where the new tnum_mul 
causes issues with its imprecision should also be possible; but as as 
Eduard suggested, this was added "just to have selftests that can detect 
the change."

---

Nandakumar