Does not run on Intel mac

[larstvei]

Hi!

Thank you for this great interface for sqlite3!

I shared a babashka script using this pod to a user running an Intel mac, resulting in the following error:

java.io.IOException: Cannot run program "/Users/user/.user/pods/repository/org.babashka/go-sqlite3/0.2.4/mac_os_x/x86_64/pod-babashka-go-sqlite3": error=86, Bad CPU type in executable

I found #12, and suggested downgrading 0.1.0, which worked perfectly. I think maybe the fix for M1 macs broke intel macs? I run an M1 mac myself, so it's hard for me to verify that the issue was introduced by #12, but it is so far my best guess!

[borkdude]

I have an Intel mac here as well (as well as an M1) one and can reproduce with 0.2.4. 0.2.3 still works for me.

[borkdude]

I found the issue, it turns out that CircleCI silently changed their runner to M1:

https://app.circleci.com/pipelines/github/babashka/pod-babashka-go-sqlite3/117/workflows/39932712-bde5-40ad-b4ff-3e5af189dd9e/jobs/191?utm_campaign=vcs-integration-link&utm_medium=referral&utm_source=github-checks-link&utm_content=summary

[borkdude]

I thought adding GOARCH=amd64 to the go invocation would fix this problem, but apparently not, with 0.2.5 I still get the same problem.

@lispyclouds perhaps you can take a look at this go stuff?

[larstvei]

I am glad that you were able to get something out of the not-very-informative bug report!

[borkdude]

At least you can still use 0.2.3 for now. Will take a look later. In can make a specific intel runner on circle to fix the problem for sure.

[lispyclouds]

I will take a closer look soon.

[lispyclouds]

Found the issue:

• The build step is correct, it does build the amd64 one as expected
• But, the test script has another build step which doesn't take in that param and overwrites the binary resulting in the amd64 becoming arm64 🤦🏽
• It seems we always were building amd64 binaries for mac and rosetta was hiding that