Juan Francisco Cantero Hurtado wrote: > On Tue, Feb 16, 2016 at 06:04:29PM -0500, Michael McConville wrote: > > Juan Francisco Cantero Hurtado wrote: > > > I've seeing a bunch of double-free and use-after-free in htop. > > > Please, run your tests with "MALLOC_OPTIONS=CFGJU htop" and fix > > > the code yourself in the upstream github repo if you can. > > > > Hm. I've been running it with a brutal malloc.conf for more than a > > day with no issues. How immediately and regularly are you seeing > > these errors? Are you running something process-intensive like dpb > > on the machine? > > Yes. Sometimes I just see the crash when I change some option. > Unfortunately, I only saved the coredump from the crash which I > reported.
I see. I take a pretty tedu approach to htop - the only command I use is 'q'. This may change at some point, but I can't be trusted to maintain the less common features. I pulled the stub code from the FreeBSD port, but in at least one case it's been insufficient and led to a null pointer deref. > > We fixed a few memleaks recently. Maybe this is related. > > Maybe. If you look the port which I sent to Pedro, I disabled the > stripping of symbols, so we could import the port after of the cvs > lock and to get better reports from the users. We can't test every > corner case in htop. > > IMHO, the port is good enough as-is now. Yeah, I wasn't too upset that it missed the 5.9 lock for this reason. Six months of -current testing will be useful.
