Configuration updates to compile with GCC 8 #2803
Conversation
|
Can one of the admins verify this patch? |
|
ok to test |
ghost
mentioned this pull request
|
Unfortunately this patch also yields a system that will not IPL, with no hostboot console output (openbmc/openbmc#3531). Reverting buildroot to GCC6 allows hostboot to start but it then immediately fails in targeting (open-power/hostboot#173) |
|
LF-Luis
force-pushed
the
master
branch
6 times, most recently
from
91ef3b3
to
28a0963
Compare
|
I attemped to build this patch on master on a RHEL 7 machine w/ witherspoon defconfig and it passed; I don't get why it's giving inconsistent results vs. local testing. |
LF-Luis
force-pushed
the
master
branch
4 times, most recently
from
95ba593
to
bd87ba8
Compare
|
retest this please |
2 similar comments
|
retest this please |
|
retest this please |
|
I'm assuming this PR is for testing only and the included patches will go to their relevant repositories instead? The Linux Makefile patch is interesting, @shenki can we handle that with a config option instead? This should probably update the other defconfigs as well, not just P9 platforms. |
|
The kernel changes are a non-starter. If we end up requiring this we will upstream them to the powerpc or top level makefile. |
LF-Luis
force-pushed
the
master
branch
2 times, most recently
from
7dff595
to
2c43c72
Compare
|
Running CI without the Kernel changes to see if new sizing still holds. |
Appears to have worked for some platforms but not others - that's 2x the normal size though I suspect something else is up there. |
|
@sammj 2x the normal size is the scripting breaking down -- if you're even 1 byte over, dd spits out another block since the block size is set to the allowed partition size. Presto, double size for a slight overage. It made debugging fun for a while until I figured out what was going on. 😄 |
|
@madscientist159 Haha, that does ring a bell. Ugh, my kingdom to find time to finish off the |
|
@shenki Can we initiate the process to pull the kernel changes to upstream. This is the only way after a month of trying we have been able to fit everything in the space constraint. Otherwise, we'll need help to somehow prune the content down or something. |
Yes. Please hit me up on slack if you need instructions on how to send kernel patches. |
LF-Luis
force-pushed
the
master
branch
6 times, most recently
from
a102326
to
daf3e72
Compare
Edit op-build config files to use GCC 8 compiler. Temporarily added 0001-Fix-gcc8-signature-validation-anomaly HB patch. Temporarily added 0002-GCC8-Code-Fixes HB patch. Has 0003-space-saving-compile-options patch to reduce size of bootkernel. Signed-off-by: Luis Fernandez <[email protected]>
For P8 configs use GCC 6, and for P9 configs use GCC 8. For P10, GCC 8 will also be selected. Signed-off-by: Luis Fernandez <[email protected]>
|
#3004 removes the P8 platforms, so that solves the P8 problem. |
|
The ci/build* scrips have changed significantly since this was first proposed, and there are plenty of other PRs (here and on Hostboot, OCC etc) proposing solutions for compilation with modern compilers.. so far, looks like the theory of 'removing power8 entirely' is winning (#3145). I'll politely close this, but if there's any aspect of this that we should still discuss, feel free to reopen/speak-up |
|
Thank you for closing this. |
No description provided.