on armv7 hardfloat CHOST, dev-lang/v8 fails to build. Or better, the QA hook at the end complains about -DUSE_EABI_HARDFLOAT=0 while -mfloat-abi=hard was set. It looks like v8_use_arm_eabi_hardfloat% in "build/common.gypi" should be set to true when hardfp is in use. Otoh, I don't know what's the best way to achieve this without hard-patching the file. Reproducible: Always
Created attachment 300703 [details, diff] force enable hardfp patch inherit flag-o-matic if is-flagq -mfloat-abi=hard; then epatch "${FILESDIR}"/v8-arm-enable-hardfp.patch fi might be enough
I think we could accomplish the same by adding this to the ebuild: export GYPFLAGS="v8_use_arm_eabi_hardfloat=true"
that is-flagq check won't work as it ignores default toolchain settings. really the code should be looking to see if __SOFTFP__ is defined. v8_use_arm_eabi_hardfloat should be used whenever that isn't defined via `${CC} ${CFLAGS} ...`.
I think the problem is that we have to pass through GYP. I'm not really a GYP nor v8 buildsystem expert. So I'm leaving the correct fix to you (sob).
I'm not really sure why the code uses a made-up macro called "USE_EABI_HARDFLOAT" instead of looking for whatever macro the compiler defines when hardfloat is enabled. Mike's (vapier) suggestion seems like the way to go.
Maybe they want to have another level of indirection of some kind? It looks quite odd anyway.
(In reply to comment #3) > that is-flagq check won't work as it ignores default toolchain settings. > really the code should be looking to see if __SOFTFP__ is defined. > v8_use_arm_eabi_hardfloat should be used whenever that isn't defined via `${CC} > ${CFLAGS} ...`. That makes sense to me. Fabio, could you please notify V8 developers about this? It really works best when person experiencing the issue reports it, and I can take it from that point and work towards an upstream fix (or downstream one, reviewed by upstream). Here's the mailing list: http://groups.google.com/group/v8-users And the bug tracker: http://code.google.com/p/v8/issues/list
@arm please get this fixed or drop your keywords.
~arm keyword dropped
Upstream bugs: http://code.google.com/p/v8/issues/detail?id=539 http://code.google.com/p/chromium/issues/detail?id=31274
(In reply to comment #10) > Upstream bugs: > http://code.google.com/p/v8/issues/detail?id=539 > http://code.google.com/p/chromium/issues/detail?id=31274 Thank you, this applies to bug #401827 though and I've made changes there. If you're interested in getting _this_ bug fixed, please report it upstream and post the link here.
follow ups are being made to Bug 429298 (in the See Also field)