We don't have the gcc builtins for atomic ops patch in the packaging either.
i grabbed the qemu-kvm source package to see which other patches we should look at:
- 1000-undo-earlier-static: seems useful for static builds; I think I send a static build fix upstream a while ago; don't think we really need this for now
- 2000-vmmouse-adapt-to-mouse-handler-changes: probably x86 specific
- arm-higher-initrd-load-address: you have that now
- arm-ignore-writes-of-perf-reg-cp15-with-crm-12: I think you have a more complete implementation
- caps-lock-key-up-event: probably useful for the SDL UI, should really go upstream
- Detect-and-use-GCC-atomic-builtins-for-locking: definitely missing in our tree
- larger_default_ram_size: bumps default RAM size, probably only applies to x86 vms, don't care
Yes, the GCC builtins stuff was sent upstream (by me) but two comments were made:
- ppc maintainer didn't want this merged as he said basically didn't trust the gcc builtins and thought it would be slower
- I think Paul Brook made the comment that the code which was being patched could go away entirely
We don't have the gcc builtins for atomic ops patch in the packaging either.
i grabbed the qemu-kvm source package to see which other patches we should look at: earlier- static: seems useful for static builds; I think I send a static build fix upstream a while ago; don't think we really need this for now adapt-to- mouse-handler- changes: probably x86 specific initrd- load-address: you have that now writes- of-perf- reg-cp15- with-crm- 12: I think you have a more complete implementation key-up- event: probably useful for the SDL UI, should really go upstream and-use- GCC-atomic- builtins- for-locking: definitely missing in our tree default_ ram_size: bumps default RAM size, probably only applies to x86 vms, don't care
- 1000-undo-
- 2000-vmmouse-
- arm-higher-
- arm-ignore-
- caps-lock-
- Detect-
- larger_
Yes, the GCC builtins stuff was sent upstream (by me) but two comments were made:
- ppc maintainer didn't want this merged as he said basically didn't trust the gcc builtins and thought it would be slower
- I think Paul Brook made the comment that the code which was being patched could go away entirely
http://<email address hidden> /msg25560. html
I'm happy if you take a fresh look at this! :-)