On 1/18/22 13:57, Andrea Bolognani wrote: > On Tue, Jan 18, 2022 at 12:30:30PM +0100, Michal Prívozník wrote: >> On 1/18/22 11:14, Andrea Bolognani wrote: >>> These changes seem to have made ASAN very unhappy, see >>> >>> https://gitlab.com/MichalPrivoznik/libvirt/-/jobs/1985244739 >>> https://gitlab.com/MichalPrivoznik/libvirt/-/jobs/1985244740 >>> >>> Tim, do you have any idea why that would be the case? My uneducated >>> guess is that the environment needed by ASAN is somehow lost when the >>> dnsmasqmock.py script is called, but I'm unfamiliar with how these >>> tools actually work. >> >> [...] I >> could mock virFindFileInPath() just like I am now, except let it return >> a predictable path (say /usr/sbin/dnsmasq) and then use >> virFileWrapper...() to redirect /usr/sbin/ to abs_srcdir. >> >> Alternative to all of this is to keed virCommandSetDryRun() just like I >> had in one of previous patches. Remind me please, what was the issue >> with that? > > The Python script approach seemed simpler, but in light of this issue > I guess that argument has gone completely out of the window :) > > Can you please try a version of this series with your original > dnsmasq mocking approach in CI and see whether ASAN is happy with it? > If so, we can just go ahead with that one. > Will do. Although, since virCommandRun() wouldn't actually execute anything I don't expect ASAN to raise any issues. Meanwhile, I'm testing the approach I've outlined: https://gitlab.com/MichalPrivoznik/libvirt/-/pipelines/450382098 Let's see how it runs. Michal
