[litmus,test] Fix a few glitches #1087
Conversation
I agree @maranget! I think Github has both Arm and x86 runners so we would be able to both compile and run tests. It might also be helpful if we could (manually) run the regression on a PR that makes changes to |
|
We could at least compile the tests.... |
|
... and we thus need either to install the right cross-compiler and tools, whatever the platform, or be able to select the appropriate platform to use the local compiler. |
|
I had a try for a periodic action that tests litmus, x86_64 only. See PR #1090. |
|
No objection from my side |
|
No objection from my side either. The only question I had is whether it’s necessary to require v8.4, for many users that would mean that the binaries wouldn’t run. |
maranget
force-pushed
the
fix-litmus-test-aarch64
branch
from
57c86f6 to
9835c8e
Compare
|
Good point @relokin. I attempted a less demanding requirement: armv8.1 with FEAT_LRCPC. Apparently it works. |
maranget
force-pushed
the
fix-litmus-test-aarch64
branch
from
9835c8e to
27ebeee
Compare
+ Fix reference to non-existing directories in some of catalogue index files. + Change Makefile variable for `make litmus-aarch64-test RUN_TESTS=true` to function. Additionaly some catalogue tests use the `LDAPR` instruction. As a consequence, gcc compilation flag -march=armv8.1-a+rcpc.
1. Less iterations for aarch64 kvm tests
2. Some tests are not terminating,
do not run their batch.
Restore
make litmus-aarch64-testfunctionality.