Yes, I agree that the defaults are in the wrong order. The platform-specific file openers should be first and the opinionated fallbacks should come later.

The fix is simple:

diff --git a/src/cmd/pprof/internal/commands/commands.go b/src/cmd/pprof/internal/commands/commands.go
index 51397a3..2c5e2d0 100644
--- a/src/cmd/pprof/internal/commands/commands.go
+++ b/src/cmd/pprof/internal/commands/commands.go
@@ -82,7 +82,8 @@ func PProf(c Completer, interactive **bool, svgpan **string) Commands {
 // browsers returns a list of commands to attempt for web visualization
 // on the current platform
 func browsers() []string {
-       cmds := []string{"chrome", "google-chrome", "firefox"}
+       var cmds []string
+
        switch runtime.GOOS {
        case "darwin":
                cmds = append(cmds, "/usr/bin/open")
@@ -91,7 +92,7 @@ func browsers() []string {
        default:
                cmds = append(cmds, "xdg-open")
        }
-       return cmds
+       return append(cmds, "chrome", "google-chrome", "firefox")
 }

Any other opinions? If ok, I will send a CL.

I suspect the fallbacks will never be used in this case (well, unless the OS opener does not exist). We could potentially get rid of them altogether?

I would keep them, since I don't know how dragonfly or plan9 implements that.

Wouldn't a preferred fallback to $BROWSER be better for those OSes?

I honestly don't know, so I suggested the minimal fix.

My guess for best order of operations would be ["os-specific opener", $BROWSER if set, "chrome", "google-chrome", "chromium" "firefox"].

The fallback list also has other problems like the "google-chrome" binary sometimes being called "google-chrome-stable", beta versions of browsers are never used, Opera and Safari aren't mentioned, etc., but those are second-order concerns.

CL https://golang.org/cl/12201 mentions this issue.