mdempsky

* Go notes

To fit into Go's current package-oriented compilation mode, we can
define a set of top-down root entry-points into each package.

Any global variable initializations and init functions are always part
of the root.

For package main, the main function is also a root.

mdempsky

I'm adding a feature to the Go compiler to make debugging it easier. That is, debugging of the Go compiler itself, not of end-user code. (I believe further details are irrelevant, but for concreteness this is CL 328909.)

This feature captures a lot of stack traces, but the expected number of unique frames across all traces is modest. For performance, I expected I would want to use runtime.Callers, and then manually process and memoize the uintptr values. I haven't had a need for runtime.Callers

mdempsky

;; Performance suggestions from
;; https://emacs-lsp.github.io/lsp-mode/page/performance/
(setq gc-cons-threshold 100000000)
(setq read-process-output-max (* 1024 1024)) ;; 1mb

(tool-bar-mode -1)
(menu-bar-mode -1)

(require 'package)

mdempsky

#!/bin/sh -e

remote=kyuubi
path=wd/go

# Make a snapshot commit of the current working tree.
git add -u
index=$(git write-tree)
commit=$(git commit-tree $index -p HEAD -m "gotry checkpoint")

mdempsky

package main

import (
	"fmt"
	"sort"

	advent "github.com/mdempsky/advent2020"
)

func main() {

mdempsky

// +build ignore

package main

import (
	"bytes"
	"fmt"
	"go/format"
	"go/types"
	"io/ioutil"

mdempsky

Language features to prototype

unsafe.Add, unsafe.Slice
Allow anything to be convertible to boolean type (zero value maps to false; non-zero maps to true)
Generalize && and || to non-boolean types (TODO: find golang proposal for this)
Ternary operator ?: (golang.org/issue/33171, and others)
Safe navigation operator ?. (golang.org/issue/42847)
Making nil a universal zero value (golang.org/issue/35966)
  - Allow T() as zero value for T? Or only allow nil to be explicitly converted to arbitrary types like T(nil)?
Make T(v) addressable (golang.org/issue/9097)

mdempsky

One of the new features in WPA3 is 256-bit AES.
E.g., see https://apps.nsa.gov/iaarchive/library/reports/wpa3-will-enhance-wi-fi-security.cfm

I'd like to enable this on my home network, which uses two Archer A7s.
(These are basically rebranded Archer C7 v5s, which Amazon was selling at a discount.)

The Archer A7 uses the "qca988x hw2.0" chipset,
which is supported by Linux's ath10k driver:
https://github.com/torvalds/linux/tree/master/drivers/net/wireless/ath/ath10k

mdempsky

### Keybase proof

I hereby claim:

  * I am mdempsky on github.
  * I am mdempsky (https://keybase.io/mdempsky) on keybase.
  * I have a public key ASD01SJVmVpvKqm1YT7Wjm3gxbKkc-pNR81jfLapyMzr6go

To claim this, I am signing this object:

mdempsky

$ go tool compile -m r.go  | grep -e leaking -e new.int
r.go:4:12: leaking param: f to result ~r1 level=1
r.go:10:10: new(int) escapes to heap
r.go:26:10: F2 new(int) does not escape


$ cat r.go
package p

//go:noinline