WFM on macos 10.14. Wondering what the aarch64 failure is about though.

tests/coalesce-tract/out/-P_--coalesce_--reorder_-z11_-Z11_-y_STATEFP10_-y_COUNTYFP10_-l_merged.json.check.out tests/coalesce-tract/out/-P_--coalesce_--reorder_-z11_-Z11_-y_STATEFP10_-y_COUNTYFP10_-l_merged.json differ: char 5105, line 34
make: *** [Makefile:94: tests/coalesce-tract/out/-P_--coalesce_--reorder_-z11_-Z11_-y_STATEFP10_-y_COUNTYFP10_-l_merged.json.check] Error 1
make: *** Waiting for unfinished jobs....

@ofborg build tippecanoe

Yup, seems to persist.

I have no idea what is going wrong on aarch64... What is better way to proceed? Mark as not supported on aarch64?

Sad, but I guess it's the best thing to do. Might be worth opening a ticket upstream letting them know that their coalesce-tract test may have issues on non-x86.

It doesn't help that:

1. They appear to be doing bytewise comparisons of json output, which will obviously be unnecessarily sensitive to e.g. object key order.
2. We only have access to ofborg for aarch64, and can't inspect its filesystem after failure.

Sad, but I guess it's the best thing to do. Might be worth opening a ticket upstream letting them know that their coalesce-tract test may have issues on non-x86.

• marked as broken on aarch64
• upstream issue coalesce-tract test fails on aarch64 platform mapbox/tippecanoe#881

@risicle I've just tried to build tippecanoe on my aarch64 machine and it works just fine. Can we unbroke it on aarch64 and perhaps disable this test on aarch64 to not fail on CI?

Having worked a bit more with aarch64 packages now, there are quite a few which have slight disagreements in float precision with "expected" results that were presumably generated on x86. I think this is due to the existence of the x87 80bit extended precision mode.

Short answer: yes, but I would raise the issue upstream too if it isn't already.

Short answer: yes, but I would raise the issue upstream too if it isn't already.

Here is upstream issue: mapbox/tippecanoe#881.