While taking a stab at building a Linaro Android NDK, I've looked at the AOSP 4.8 toolchain being built.
Some of the stuff there is needed for NDK compilers (so we should merge it in order to reduce the number of patches we have to apply in individual builds), some stuff looks like a good idea, some stuff looks like a good idea but doesn't affect ARM, some stuff is bogus but possibly needed (forcing Androidisms without taking care of other platforms), some stuff is not very important.
The ones I need apply on top of Linaro GCC 4.8 without changes.
I'd categorize them as:
Merge:
0002-4.8-Fix-crash-on-null-context-in-cp-mangle.c-write_u.patch
0008-4.8-Handle-NULL-return-value-of-setlocale-in-libstdc.patch
0013-4.8-Fix-ARM-GCC-fail-to-links-code-using-64-bit-atom.patch
0014-4.8-Patch-GCC-libstdc-to-work-with-Clang-in-C-11.patch
0015-4.8-Enable-_GLIBCXX_HAS_GTHREADS-in-GCC-4.8.patch
0016-4.8-Support-OpenMP.patch
0019-4.8-Add-mclinker-support-fuse-ld-mcld.patch
0021-4.8-Fixed-GCC-4.8-ICE-gen_thumb_movhi_clobber-at-con.patch
0023-4.8-OpenMP-Better-CPU-count-detection-for-Linux.patch
0024-4.8-Fix-openmp-support-to-use-libgomp-config-linux.patch
0025-4.8-Use-READELF_FOR_TARGET-in-GCC-4.8.patch
0026-4.8-GNU-libstdc-enable-linux-futex.patch
0028-Misc-fixes.patch (better yet, fix the patches needing this fixup)
0029-Add-fuse-ld-mcld.patch (merge with 0019-4.8-Add-mclinker-support-fuse-ld-mcld.patch)
0032-Fix-GNU4.7-4.8-libstdc-for-clang.patch
0036-4.6-4.7-4.8-Backport-of-the-patch-from-trunk-4.8-for.patch
0039-4.8-Fix-ICE-on-g-.dg-cpp0x-lambda-lambda-defarg3.C.patch
0040-Add-an-internal-switch-minline-thumb1-jumptable-to-A.patch
Nice to have, but not very relevant:
0044-4.8-Backport-diagnostic-color-support-from-trunk.patch
Don't merge:
0001-4.8-Disable-warning-the-mangling-of-va_list-has-chan.patch
0003-4.8-Workaround-for-Android-__cxa_finalize-bug.patch (That __cxa_finalize bug has been fixed in current Android anyway)
0007-4.8-Simplify-GCC-4.7-BASE-VER.patch
0042-4.8-Simplify-GCC-4.8-BASE-VER.patch
Needs further investigation/fixing:
0004-4.8-Enable-assembler-linker-default-for-security.patch
0005-4.8-Enable-x86-arm-gcc-defaults-Remove-march-mtune-m.patch
0010-4.8-Enable-armv7-thumb2-multilib-for-arm-eabi.patch
0011-4.8-Backport-a-gimplify-fix.patch (seems bogus, since 55245 is fixed)
0017-4.8-Disable-libstdc-versioning.patch (needs to be made android-only)
0022-4.8-Use-default-visibility-for-the-__cxa-functions-i.patch
0027-4.8-Maintain-default-visibility-for-helper-functions.patch (needs to be made android-only, see my replacement patches)
0030-Disable-aarch64-linux-unwind.h-for-now.patch
0034-Fix-ARM-GCC-4.7-4.8-generates-insufficient-alignment.patch (see discussion on bug report, but we may need this to remain ABI compatible with AOSP)
0041-4.6-4.8-Add-stack-unwinding-directives-for-__aeabi_i.patch
0043-4.8-Always-enable-eh-frame-hdr-for-static-executable.patch
Not relevant for ARM platforms (therefore didn't bother to investigate further):
0006-4.8-Port-MIPS-Android-support-to-GCC-4.8.patch
0009-4.8-Add-mstack-protector-guard-to-x86-compilers.patch
0012-4.8-Enable-MIPS-floating-point-madd-msub-nmadd-nmsub.patch
0018-4.8-Use-memalign-instead-of-posix_memalign-in-GCC-x8.patch
0020-4.8-Fix-MIPS-GCC-to-emit-m-elf32ltsmip-for-linker.patch
0031-Update-mstack-protector-guard-support-based-on-upstr.patch
0033-Add-new-GCC-MIPS-option-mldc1-sdc1.patch
0035-Remove-hardcoded-m32-in-compiler-options.patch
0037-4.7-4.8-Release-basic-tuning-for-new-Silvermont-arch.patch
0038-4.7-4.8-Extended-Silvermont-tuning.patch
0027-4.