Thanks for reporting. It worked perfectly on x86_64 when I tried it. Will look into the issue and test on another installation.

Updated the Julia package. Tested on two x86_64 systems.

Does it work for you as well now?

Julia working fine now, thanks.

Great! Closing this one.

Apparently you fixed the problem by not shipping the sysimg.so file, which makes Julia's startup slower. I'd just wanted to note that I've experienced the same problem when making a Fedora RPM, and I've filed an issue. The solution would be to create an image suited to all CPUs. https://github.com/JuliaLang/julia/issues/6715

Thanks for the information. Hopefully the Julia developers will find a good solution for this. As you are probably aware, Arch Linux tries to avoid patching and tries to ship a package that is as close to upstream as possible. Unless I have misunderstood, this is an upstream problem and not a packaging problem.

Closing this one as "upstream". If you think this is a packaging problem, please re-open this issue and I'll do my best to look into it. Thanks.

The workaround solved the problem, with the tradeoff of making the startup slower.

 FS#39721  suggested that this problem might be solved, so I reopened this bug while the issue is looked into.

Now works without removing sys.so. Will try with the latest commit as well.

Works without sys.so. Updating the package in community and closing this bug.

Tested on another computer, and sys.so was creating problems there. Removing sys.so again.

Do you have any details on the machine where the problem happened? Are you setting JULIA_CPU_TARGET=core2? Please report issues upstream so that we can try to improve things.

Yes, JULIA_CPU_TARGET is set to core2.

Vesa Kärkkäinen, please report this bug upstream if it is still a problem and if it doesn't work with the latest julia package (version 0.3.0, the release version).

Related:  FS#41666 .

I've opened an upstream bug: https://github.com/JuliaLang/julia/issues/8119

Alexander Rødseth: You're setting JULIA_CPU_TARGET=core2, but apparently you are explicitly removing the system image after the build. This shouldn't be needed. See the discussion upstream.

Milan Bouchet-Valat, that is incorrect. sys.so is only removed if julia is broken and does not start: https://projects.archlinux.org/svntogit/community.git/tree/trunk/sysfix.install?h=packages/julia

Yes, but I'd argue it's not a good strategy : better get a failure so that it can be reported clearly and fixed; else nobody sees any error message, and users think Julia or the Arch package are necessarily slow to start. We're trying to debug this upstream.

It appears that JULIA_CPU_TARGET=core2 should rather be MARCH=${CARCH/_/-}

Updated the PKGBUILD and rebuilding the julia package. The updated package will appear in [community] shortly.

Milan Bouchet-Valat, if it is not a good packaging strategy, why is Julia asking users/packagers to remove sys.so if it's not working, instead of crashing hard, outputting debug information and asking users to file a bug report? Julia says: "Target architecture mismatch. Please delete or regenerate sys.{so,dll,dylib}."

Well, I guess this message is intended at users who used "native", not at distributors who normally use baseline architectures which don't fail anywhere (as you and I do). And it says "delete or *regenerate*".

Anyway, glad to know it works now! Please feel free to report bugs upstream earlier if you find problems like this in the future (including to fix error messages).

Will do. I'm also glad it's working now. :)

Thanks for caring about julia and the julia package. Closing this bug.