incorrectly identified arch as "darwin-ppc" or "darwin-i386" on macOS arm64

[BenWibking]

Describe the bug

Issue 1. When packaging the Visit install on a macOS arm64 machine, build_visit labels the arch as darwin-ppc.

Issue 2. When running the launcher script $VISIT_INSTALL_DIR/bin/visit, it tries to load the binaries from a darwin-i386 subdirectory.

To Reproduce

Steps to reproduce the behavior. For example:

1. Run ./build_visit on latest develop on a Mac
2. See the directories and files created by the packaging script are labeled darwin-ppc.

Expected behavior

It should identify the arch as darwin-arm64 (or similar).

Attachments

Build log:

Desktop

• OS and version: macOS Sequoia 15.1
• VisIt Version: develop
• Server info (if applicable):

Additional context

For issue 2, it looks like the problem is in internallauncher:

        elif self.os == "darwin":
            mach = self.uname("-m")
            release = self.uname("-r")
            version = [int(x) for x in release.split(".")]
            self.osver = version[0]
            if version[0] >= 11:
                if mach == "x86_64":
                    supportedarches.append("darwin-x86_64")
                    supportedarches.append("darwin-i386")
                else:
                    supportedarches.append("darwin-i386") # this line gets executed

The value of mach on this machine is arm64.

[BenWibking]

If internallauncher is updated to include a mach == "arm64" case, I think this issue should also get fixed: #18867.

[cyrush]

@BenWibking thanks! We will try to resolve this during our release process for 3.4.2.

[markcmiller86]

Ok, so this is really just a problem with internallauncher find the install. It isn't something about how masonry or build_visit work to build and deploy VisIt for M1.

[markcmiller86]

Ok, I have a full native arm64 build of VisIt on my M1. But, packaging is failing at some point. I am seeing vestiges of what is reported here...

[sce2aux:~/visit/3.4RC/release] miller86% find . -type d -name darwin-ppc
./build-mb-3.4.2-darwin-23-arm64-release/build.release/_CPack_Packages/Darwin/TGZ/visit3_4_2.darwin-ppc/3.4.2/darwin-ppc
./build-mb-3.4.2-darwin-23-arm64-release/build.release/_CPack_Packages/Darwin/Bundle/VisIt-3.4.2/VisIt.app/Contents/Resources/3.4.2/darwin-ppc
./build-mb-3.4.2-darwin-23-arm64-release/install.release/3.4.2/darwin-ppc

I have to discover which piece of software is responsible for generating that darwin-ppc name.

[markcmiller86]

Ok, this rabbit hole is a bit deeper now. None of the codesigning and notarization steps for arm64 macs are working in my release builds because somewhere, its decided the incorrect archicture moniker.

[markcmiller86]

Turns out there were a lot of cases where darwin-<arch> is handled...

1596180