.. moving to ports@ On 2018/05/16 11:32, Marc Espie wrote: > On Wed, May 16, 2018 at 08:41:17AM +0000, Stuart Henderson wrote: > > On 2018-05-16, William Orr <[email protected]> wrote: > > > Clicking the password field will consistently cause that tab in firefox > > > to crash with a pledge violation (calling fork): > > > > > > firefox[75379]: pledge "proc", syscall 2 > > > firefox[99617]: pledge "proc", syscall 2 > > > firefox[89996]: pledge "proc", syscall 2 > > > firefox[29564]: pledge "proc", syscall 2 > > > firefox[58111]: pledge "proc", syscall 2 > > > firefox[97980]: pledge "proc", syscall 2 > > > firefox[37363]: pledge "proc", syscall 2 > > > > > > Is anyone else seeing something similar? I've repro'd this in safe mode > > > with add-ons disabled. I'm runnning a snapshot as of 3 days ago with > > > firefox from packages. > > > > > > % pkg_info firefox > > > Information for inst:firefox-60.0 > > > > > > Following is a full dmesg. Let me know if there's other info that I can > > > provide. There are other firefox pledge violations in there, but I have > > > no indication that they're related. > > > > The Firefox port currently includes some experimental pledge code, > > see https://marc.info/?l=openbsd-ports&m=152623658627250&w=2 for > > information about debugging and a way to disable it without > > recompiling. > > > > From what I've seen, I'm reasonably sure the experiment should be turned > off for now, while landry fixes the most obvious bad cases that various > people have reported, then turn it back on for finer-grained issue. > > More reports of the same thing are useless/counter-productive. We're > reaching the point of diminishing returns where it takes longer to answer > emails/classify failures into "already known/new"... > ... which is a recipe for issues to fall between the cracks, because people > WON'T report new issues, or they will be dismissed as the same as something > that's already known...
Yes, I agree. I think we should at least add the known-needed ones like getpw (even if we don't want them long-term). We have no hope to analyse the code and figure out a complete list of what's actually needed as is usual with pledge.. > BTW, if you're supposed to start dbus BEFORE firefox, that's cool, but then > the firefox code should be tweaked to display "please start dbus" instead > of the "helpful" error message "firefox crashed in a pledge violation (proc)". Simplest method is a wrapper script I think.
