(Apparently forgot to send it yesterday, so sending it with a small addedum.) On Tuesday, 17 March 2020 18:09:04 CET Richard W.M. Jones wrote:
My only worry about this patch is that it relies on fileName now possibly being NULL, which means if there is any case that you've missed now — or one added in future — which doesn't consider that fileName might be NULL then it'll crash (libvirtd? or virsh? I'm not sure).
In case now (even in v2) fileName is used without checking, it will crash libvirt, as the esx/vmware drivers are built-in in the library.
I wonder if therefore it would be safer to set the string to a known-good non-NULL value such as ‘strdup ("emptyBackingString")’?
Thought about that, however "emptyBackingString" seems like a magic identifier, and it would be trading one hack with another, somehow. -- Pino Toscano