Some are not. You can try this:

    diff --git a/src/runtime/mgc.go b/src/runtime/mgc.go
    index a2b6b979c1..d2f2852294 100644
    --- a/src/runtime/mgc.go
    +++ b/src/runtime/mgc.go
    @@ -144,7 +144,7 @@ const (
     
      // debugScanConservative enables debug logging for stack
      // frames that are scanned conservatively.
    - debugScanConservative = false
    + debugScanConservative = true
     
      // sweepMinHeapDistance is a lower bound on the heap distance
      // (in bytes) reserved for concurrent sweeping between GC

And observe that Go scans a few stack frames conservatively. I think it only does this if a stack frame is preempted. Most frames are scanned precisely.

As the Fine Article mentions, the JavaScriptCore GC is conservative, and V8 is considering a switch.

Someone reading your sentence might be at risk of conflating a minority position with a fringe one. Clearly this isn't the case here.

I said minority, not fringe.

But fringe isn’t far off. Among those who write GCs professionally, I was definitely on the fringe as an advocate for conservative-on-the-stack.

(I wrote most of JSC’s GC and I was one of the folks pushing for it to remain conservative at a time when V8 was accurate and SpiderMonkey transitioned to being accurate. I like that folks are now acknowledging the good choice we made in JSC, but it was a fringe/minority choice at the time for sure, and it’s still a minority choice among high performance GCs.)

I would say you're being a bit modest! JSC is deployed on hundreds of millions of devices, and if V8 does adopt a conservative GC strategy, that would be a comfortable majority of JavaScript engines making use of that approach.

It goes to show that popularity is a poor measure of technical merit, and the fringe doesn't always stay that way.

s/accurate/precise/

No.

The literature uses “Accurate GC”, “Precise GC”, and “Exact GC” interchangeably.

Famous paper on this that uses “accurate”: https://citeseerx.ist.psu.edu/document?repid=rep1&type=pdf&d...

Paper that calls it “exact”: https://dl.acm.org/doi/10.1145/2660193.2660198