Date: Tue, 5 May 2020 08:30:55 +0200 (CEST) From: Paul FLOYD <pjfloyd@wanadoo.fr> To: freebsd-hackers@freebsd.org Subject: Call for testers - Valgrind Message-ID: <453259393.837.1588660255688.JavaMail.www@wwinf1m04>
next in thread | raw e-mail | index | archive | help
Hi I've done a lot of work since the start of the year trying to reboot Valgrind on FreeBSD. Whilst there is still a way to go, the results are looking promising. I'd like to get some feedback, in particular from developers that use debug versions of Qt and libc and who are capable of telling whether any errors produced are false positives or not. All of the items in the FreeBSD bugzilla concerning Valgrind have been addressed, even the enhancement requests. Supported Platforms ~~~~~~~~~~~~~ I've set up VirtualBox VMs as far back as FreeBSD 10 and tested that at least everything builds. I've been doing fairly extensive testing (with Valgrind's own regression suite) on 12.1 amd64 and i386. I'll set up a 13.0-CURRENT VM soon to ensure that I'm ready for the next release. Hardware other than amd64 and i386 is currently not supported with FreeBSD. Porting to other platforms would be a major endeavour. Known Limitations ~~~~~~~~~~~ 1. Signal handling is the current weakspot, particularly on i386. Asynchronous are not correctly delivered to guest signal handlers if they exist. 2. DRD does not work with dlopen() 3. Massif on i386 does not work with --pages-as-heap 4. There are still numerous missing syscalls. 5. ioctl support is limited to those that encode read and write information in the ioctl value. That's the case for most ioctls but not all. 6. _umtx_op ROBUST_LISTS are not supported 7. Valgrind does not work well with capsicum enabled applications, and massif won't work at all. My github repo is here https://github.com/paulfloyd/freebsd_valgrind A+ Paul
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?453259393.837.1588660255688.JavaMail.www>