Bump BR for gcc 8.3 support; point submodule at upstream #25
Conversation
dev to master firesim 1.5.0
davidbiancolin
assigned davidbiancolin, NathanTP and zhemao and unassigned davidbiancolin
doesn't work though. This commit: builds br img unconditionally, and always adds the /sbin/init -> /init symlink for all initramfs builds.
The special /init script is needed to pipe stdin/stdout/stderr to /dev/console before devtmpfs is mounted.
accident on earlier versions of buildroot, now we do it explicitly.
|
I fixed the initramfs thing. full_test.sh sort of passes (modulo the issues in #26 which are unrelated to this PR). I didn't test on firesim itself, just disk and initramfs based builds on spike and qemu. |
There was a problem hiding this comment.
LGTM. Tested on spike and Qemu for both disk and initramfs builds, buildroot and fedora. full_test.sh passes except for some unrelated known issues (I manually tested those).
One thing to note is that I defaulted to rebuilding buildroot-based workloads all the way from the base distro every time (will be fixed eventually, tracked at #21). This used to default to /not/ rebuild, I figure this is safer for now (especially since this PR changes some buildroot details).
|
Thanks for bringing this home Nathan. |
|
This should probably merge against dev instead of master. I haven't really thought it out, but I suppose it makes sense to keep master in sync with firesim master. |
|
You're right, I carelessly opened it up against the wrong base branch. |
See firesim/firesim#290.