On Thu, 31 Mar 2022 at 18:54, Richard Henderson <richard.hender...@linaro.org> wrote: > > On 3/31/22 04:46, Peter Maydell wrote: > > FWIW, the Coverity issues are CID 1487134, 1487139, 1487151, 1487184, > > where for some reason it thinks that floatx80_compare() and > > floatx80_compare_quiet() can return 3 and thus that there is a > > potential array overrun. (I've marked these all as false positives > > in the UI, anyway.) > > Interesting about '3'. I'll have a look.
Unfortunately it doesn't seem to give its reasoning for deciding that the function can return [-1..3] rather than [-1..2]. But maybe it will make more sense to you. PS: while you're there, there are also a bunch of new TCG related issues where it alleges array indexes being out of bounds. I suspect these are false positives, but it's probably faster for you to analyse them. (I have a feeling Coverity can get confused and claim an error because it's looking at an array size it has cached from one target's NB_MMU_MODES value and a code flow for a different target with a different NB_MMU_MODES.) -- PMM
