On Mon, Jul 11, 2011, Olivier Martin wrote:
The Tianocore/EDK2 tree is divided in different packages. Each package has its own maintainer. The patches needs to be reviewed by the maintainer before to be committed.
What's the best way forward here? Opening a bug with each package and/or mailing each maintainer, or perhaps there's a list I should start a discussion on? My patches aren't really patches, they just paper over the real issues at this point, but I'd rather hear from the maintainers before proposing this or that approach. If other maintainers are mostly working on other arches, like x86, am I supposed to test whether the issue affects x86 and that the patch doesn't regress x86?
I cannot guarantee all these patches will be committed to the repository soon because I do not have control on these packages. But some of these patches are in process to be pushed upstream.
That's understandable; it does bet for the question of staging changes; I see you currently maintain this patch that one applies before build; it's not ideal though because it shows as changes in the local checkout. Maybe we should be recommending the use of git-svn or bzr-svn and hosting both a clone of upstream + a modified branch with ARM/Linaro patches already applied on top?
This error is due to the UEFI binary used some relocation symbols not supported by Tianocore. It could be another Tianocore bug related to the use of a newer GCC.
Ok; I need to try with the toolchain you linked to too.
Our toolchain isn't ideally suited to build UEFI as it targets linux, but I believe we should fix issues in building UEFI with such a setup (one can build linux, barebox and u-boot fine with the Linaro linux cross toolchains). The other big difference outside of GCC versions is the default setup of our toolchain: it defaults to armv7-a and thumb.