On 14 November 2011 15:06, Avi Kivity <a...@redhat.com> wrote: > On 11/14/2011 05:00 PM, Peter Maydell wrote: >> On 14 November 2011 14:59, Peter Maydell <peter.mayd...@linaro.org> wrote: >> > It would be good to see some testing that this works on all >> > our random oddball hosts (MacOSX, for instance, which doesn't >> > use GNU ld) and architectures we apply this... >> >> *before we apply this... > > We're unlikely to see testing before we apply the patch, and for oddball > archs, even afterwards.
Yes, but if you put this change in just before release you get much less testing than if you do it with several months before release. This kind of change makes me nervous. > What we can do the qualify it on a build test > (and assume that if it builds, it runs, which I think is a safe assumption). One of the failure cases I was thinking of is that if PIE means the platform's loader puts things in a different bit of the address space this might break TCG's assumptions about maximum distances between the codegen buffer and host C code. (That's a bug in TCG really but it would still be near-to-release breakage.) But yes, a build test would be a good start. (My money's on it failing to build on MacOSX.) PS: what's the -DPIE needed for? -- PMM
