When merging v8 (required e.g. by chromium) the compilation is aborted with the error message: ERROR: Binary compiled with -mfloat-abi=hard but without -DUSE_EABI_HARDFLOAT
Created attachment 319940 [details] emerge --info
Created attachment 319942 [details] Build log
Created attachment 319944 [details] emerge -pqv
Created attachment 319946 [details] Environment file
Created attachment 319974 [details, diff] patch to v8 makefile to add access to -Dv8_use_arm_eabi_hardfloat option This patch can be applied to the wrapper makefile around gyp in the v8 source tree to allow access to the -Dv8_use_arm_eabi_hardfloat option, which enables the hardfloat build for v8. This patch should be performed from the src_prepare ebuild function.
Created attachment 319978 [details, diff] ebuild patch Proposed patch to the ebuild to apply the makefile patch and enable building with the hardfp option. STATUS: Compilation works. 13 of the around 8500 automatic test cases fail. Usability not yet confirmed.
dev-lang/v8 is not even keyworded ~arm... Please let chromium@ review any ebuild changes before committing them. Generally this should be fixed upstream.
Hardfloat is now considered the default for ARM (at least armv7) so this patch is very handy. If you need access to an ARM box to test on Pawel, let me know, I can give you access to a quad core 1ghz currently running hardened hardfloat. It's current use is for blueness to do arch testing/keywording/stabling. Getting chromium working on ARM would be nice as currently webkit-gtk won't build, it keeps trying to mmap 1109438208 bytes for an output file. Firefox is in a similar situation.
Please submit the v8 patch upstream and post the link here. See <http://code.google.com/p/v8/wiki/Contributing> for more info. After the code lands upstream, we can backport it for older v8 versions in portage.
(In reply to comment #9) What are you actually after here? Auto-detection of hardfp in the v8 build system?
(In reply to comment #10) Never mind, I missed the Makefile patch. :)
(In reply to comment #10) > What are you actually after here? Auto-detection of hardfp in the v8 build > system? Auto-detection would be non-trivial, see bug #401835 . My main point is to submit the Makefile patch upstream. It looks good to me.
Fixed in CVS, thanks. Upstream review: http://codereview.chromium.org/10889026/