On 4/2/26 9:50 AM, Renaud Allard wrote: > I fuzzed for 5 minutes with random \v inputs and UBSan on both > OpenBSD 7.9/amd64 and 7.9/i386: 0 crashes out of ~55000 inputs. > > So it looks like that was the right solution. > This diff looks reasonable to me at a first glance, but I'm not comfortable to OK it without investing a lot more time in understanding its inner workings. I'm not sure if I have the spare resources for that now, so if someone else feels like picking this up, please do so.
It might also be a good idea to add a regression test with a bunch of well-chosen in-/outputs to make sure it not only doesn't crash, but also returns the desired output; Else we could as well fix it by turning it into a read->write loop. martijn@
