Older u-boot versions hardcode the use of a host gcc compiler, and gcc relies on having gperf available so we must build and install it first.
Details
Diff Detail
- Repository
- rS FreeBSD src repository - subversion
- Lint
Lint Skipped - Unit
Tests Skipped
Event Timeline
release/arm/release.sh | ||
---|---|---|
133 ↗ | (On Diff #2748) | I've tried this exact change, which sadly does not work. Discussed in IRC with imp@, building gcc(1) should not be necessary with the sysutils/u-boot-* ports. |
I've applied this change to the set of builds that are in-flight now, though I'd still prefer using the ports ian@ worked on, and avoid building gcc(1) within the build chroot entirely.
I agree moving to the ports is the right path forward, but hopefully this will get things moving again until that switchover.
release/arm/release.sh | ||
---|---|---|
130 ↗ | (On Diff #2822) | while this will work, you should transition over to using the u-boot ports for the supported boards, which pulls in the right u-boot and cross compiler... |
release/arm/release.sh | ||
---|---|---|
132–133 ↗ | (On Diff #2831) | Without splitting obj out I get "Warning: Object directory not changed from original /usr/src/gnu/usr.bin/gperf" - i.e., gperf ended up built in the src tree instead of under /usr/obj. |
release/arm/release.sh | ||
---|---|---|
132–133 ↗ | (On Diff #2831) | I don't expect this to make a difference, to be honest. When originally trying to build gperf(1) standalone, I did 'make -C /usr/src/gnu/usr.bin/gperf obj depend all install', which resulted in the same failures I've been seeing, which means something is fundamentally broken here. Splitting up the make(1) targets should not be a fix. |