redis: only link libatomic on i386

[dgilman]

Closes: https://trac.macports.org/ticket/65311

Description

This will at least stop the bleeding but should hopefully resurrect builds on i386 too.

Type(s)

• bugfix
• enhancement
• security fix

Tested on

macOS 10.15.7 19H1922 x86_64
Xcode 12.4 12D4e

Verification

Have you

• followed our Commit Message Guidelines?
• squashed and minimized your commits?
• checked that there aren't other open pull requests for the same change?
• referenced existing tickets on Trac with full URL?
• checked your Portfile with port lint --nitpick?
• tried existing tests with sudo port test?
• tried a full install with sudo port -vst install?
• tested basic functionality of all binary files?

[ryandesign]

This will not be correct for universal builds. Use the facilities of the muniversal portgroup to add the -latomic flag only for the i386 part of universal builds.

[mascguy]

Apart from the outstanding comment from @ryandesign re: Universal builds, this fixes the build issues for earlier macOS releases. (Haven't tested across-the-board, but can confirm for 10.12 through 10.14.)

[kencu]

clang doesn’t use libatomic. It’s a gcc-only library.

clang on i386 uses builtins in compiler_rt. On x86_64 it uses processor intrinsics.

[kencu]

ie this whole fix, adding libatomic to fix the i386 build with clang, is not the right fix.

[dgilman]

I don't have the hardware to test any of this stuff out so I've been flying blind. I will back out the change then. The redis makefile jumps through a bunch of hoops and needs a bunch of environment variable hand holding so I suspect that even with the hardware to test it on this is going to be an involved change.

[mascguy]

I don't have the hardware to test any of this stuff out so I've been flying blind. I will back out the change then. The redis makefile jumps through a bunch of hoops and needs a bunch of environment variable hand holding so I suspect that even with the hardware to test it on this is going to be an involved change.

@barracuda156 Sergey, would you be interested in helping to test a Universal build for this port?

[mascguy]

Given that users need this fix ASAP, let's merge.

@barracuda156 Sergey, we can fix Universal via a subsequent PR if necessary, depending on your results. Let us know, and we'll go from there!

[barracuda156]

Given that users need this fix ASAP, let's merge.

@barracuda156 Sergey, we can fix Universal via a subsequent PR if necessary, depending on your results. Let us know, and we'll go from there!

@mascguy Sure, I will save this PR link and test +universal on PPC machines, but please allow me to postpone that until the end of month! I am on a trip, and have no access to my PPC hardware at the moment. Should be back by July 1.