ffi: port semi-colon fix for riscv64 (and others)

[sxa]

This PR pulls across the fix which has been accepted upstream at libffi/libffi#968 and will therefore be in any future versions of libffi that we take.

@ShogunPanda I'm not sure if we have implemented any specific rules/restrictions on taking fixes like this so hopefully this is ok to go in as a patch since I had it merged upstream before raising this.

Fixes nodejs/unofficial-builds#235 which has the explanation of when and why it occurs.

While this includes all platforms which would have the same issue I'll note that loongarch does not show the problem in the unofficial builds on that platform but that is likely because it is using gcc and node clang. Also noting that we have loongarch64 in this PR as opposed to the upstream loongarch directory - that is because of this upstream change where the directory was renamed.

FYI @nodejs/platform-loong64 @nodejs/platform-riscv64

[nodejs-github-bot]

Review requested:

• @nodejs/ffi
• @nodejs/security-wg

[ShogunPanda]

@sxa I'm not sure about this. I guess we should wait for a official libffi fix and then run the update, or am I wrong?
FWIW, I don't mind accepting this, you did an amazing job following it up. I just want to follow the right procedure.

[sxa]

@sxa I'm not sure about this. I guess we should wait for a official libffi fix and then run the update, or am I wrong? FWIW, I don't mind accepting this, you did an amazing job following it up. I just want to follow the right procedure.

Yeah I understand that although I'm not sure when that might be. The last release (the one we're on) was in August last year. Given that it causes a build break at the moment with libffi enabled (albeit currently on an unofficial platform) I figured it would be preferable to float the patch for now and then pick up the fix officially when there is another release.

It will be included in 3.6.0 but it currently doesn't have a target release date. (Maybe @atgreen has that information as I don't think there's a regular release schedule)

[sxa]

Re-running Test Shared libraries / x86_64-darwin: with shared libraries which failed on first attempt. EDIT: Re-run looks good

[mcollina]

lgtm

I think we should follow the same pattern we have for floating patches on dependencies. I'm not overly concerned because this is still experimental and libffi is not a fast moving target.

[nodejs-github-bot]

CI: https://ci.nodejs.org/job/node-test-pull-request/73980/

[ShogunPanda]

lgtm

I think we should follow the same pattern we have for floating patches on dependencies. I'm not overly concerned because this is still experimental and libffi is not a fast moving target.

Ok, then it LGTM as well.

[nodejs-github-bot]

CI: https://ci.nodejs.org/job/node-test-pull-request/73991/