On 2025-05-05, Vagrant Cascadian wrote: > On 2025-05-05, Vagrant Cascadian wrote: >> On 2025-05-05, Efraim Flashner wrote: >>> I was trying to deploy a new generation on my pine64 and the >>> arm-trusted-firmware failed to build. I tried changing the make-flags >>> on arm-trusted-firmware to set CFLAGS=-Os but the overflow was still 216 >>> bytes. Changing it to -O2 made it 4312 bytes. Adding DEBUG=0 didn't >>> make a difference either. > ... >>> ld: address 0x500d8 of /tmp/guix-build-arm-trusted-firmware-sun50i-a64-2.12.1.drv-0/source/build/sun50i_a64/debug/bl31/bl31.elf section `.data' is not within region `RAM' >>> ld: BL31 image has exceeded its limit. >>> ld: region `RAM' overflowed by 216 bytes >>> collect2: error: ld returned 1 exit status >>> make: *** [Makefile:1557: /tmp/guix-build-arm-trusted-firmware-sun50i-a64-2.12.1.drv-0/source/build/sun50i_a64/debug/bl31/bl31.elf] Error 1 > ... >> Might be worth bisecting to see where it started failing (presuming it >> ever built natively)? I might give that a whirl... > > Well, the issue was introduced in the commit where 2.12.1 was added, > 3aed6ee07bd0d22cd6cf897eada7ea3b95f6adf5, the immediately preceeding > commit still built successfully. And bisected upstream with: guix build --with-commit=arm-trusted-firmware-sun50i-a64=$(git show | awk '/^commit/{print $2}'}) arm-trusted-firmware-sun50i-a64 git bisect start # status: waiting for both good and bad commits # good: [4ec2948fe3f65dba2f19e691e702f7de2949179c] Merge "docs(changelog): changelog for v2.12 release" into integration git bisect good 4ec2948fe3f65dba2f19e691e702f7de2949179c # status: waiting for bad commit, 1 good commit known # bad: [8cf9edba5cc3ec11ed5463f206aa5819f7fdbade] docs(changelog): changelog for lts-v2.12.1 release git bisect bad 8cf9edba5cc3ec11ed5463f206aa5819f7fdbade # good: [16b87247ed03fbe98c5c4a96c4dd00844ae53284] fix(security): add CVE-2024-7881 mitigation to Cortex-X3 git bisect good 16b87247ed03fbe98c5c4a96c4dd00844ae53284 # bad: [7e41b706e97c0e36e19d51b6a1785280e9ff4108] fix(cpus): workaround for Neoverse-N2 erratum 3701773 git bisect bad 7e41b706e97c0e36e19d51b6a1785280e9ff4108 # bad: [24526273fc504500c36029fa8c188f4e394f4f07] fix(cpus): workaround for Cortex-A720 erratum 3699561 git bisect bad 24526273fc504500c36029fa8c188f4e394f4f07 # bad: [7e4bf042a0dda901b9e5a7f79d37a34d81614e79] chore(cpus): fix incorrect header macro git bisect bad 7e4bf042a0dda901b9e5a7f79d37a34d81614e79 # bad: [9427c061eb8dee00c32d72c20d6bc231002a575c] fix(security): apply SMCCC_ARCH_WORKAROUND_4 to affected cpus git bisect bad 9427c061eb8dee00c32d72c20d6bc231002a575c # good: [bea64fd5272deaf53111135baf36abb47b73b93e] fix(security): add support in cpu_ops for CVE-2024-7881 git bisect good bea64fd5272deaf53111135baf36abb47b73b93e # first bad commit: [9427c061eb8dee00c32d72c20d6bc231002a575c] fix(security): apply SMCCC_ARCH_WORKAROUND_4 to affected cpus Yay guix build --with-commit!!!! Last time I built this version on Debian it worked fine ... but now you have me wanting to double-check! Differences in the toolchains between Debian and guix might be a clue... live well, vagrant