Memory corruption crashes every few days

[joeshaw]

Describe the bug
I have been seeing SwiftBar crashes (either SIGSEGV or SIGABRT) due to memory corruption.

I haven't been able to detect a consistent pattern with the crashes. They seem to happen every couple of days, typically after I've put my machine to sleep. They seem to happen when my machine does a macOS "power nap" wakeup, though I am not certain on that.

I'm attaching 3 crash reports to the bug. Two are segfaults (SIGSEGV) and one is a SIGABRT triggered by malloc() detecting heap corruption.

Environment:

• macOS version: 12.3.1 on Intel
• SwiftBar version: 1.4.3, installed from the homebrew cask

Additional Context:

I have 3 scripts running:

• A bash script that runs every 5 seconds
• A Go program that runs every 5 seconds
• A Go program that runs every 30 seconds

I do run Bartender. I have not (yet) tried running without it, though I previously ran earlier 1.4.x versions under it without issue.

I've attached the crash reports generated my macOS below. GitHub did not like the original ips extension, so I appended .txt to them. They are JSON files, however, but can be rendered more prettily by Console.app.

SwiftBar-2022-05-05-191332.ips.txt
SwiftBar-2022-05-06-165801.ips.txt
SwiftBar-2022-05-11-205124.ips.txt

[gimbo]

As a curious user considering migrating from xbar, can I ask: when it crashes, do the icons disappear from your menubar (i.e. SwiftBar is completely gone/stopped) or is it just that it stops updating?

I'm asking because I'm motivated to move by xbar issue #714 ("Crashing after a large number of refreshes") and there's no point migrating if the same behaviour occurs here!

I'd be surprised if it is the same error, but wanted to check. 😄

[joeshaw]

@gimbo It crashes and the items leave the menubar.

An older version (1.4.1 iirc) had an issue where items stopped updating, but that was fixed and I haven't had any problems with updating since.

[melonamin]

I'm trying to catch this in a debug session, I've been running 3 simple plugins with 5s refresh period for a week and a half now... no luck.

I'll try to have more complex plugins and somehow trigger sleep-wake. Stay tuned...

[melonamin]

@joeshaw can you share with me any of your plugins?

So far SwiftBar is solid for me, with two caveats:

• I'm running an unrelease version(but I don't think it matters)
• I'm on ARM

[bobobox]

Similar experience here - two SIGSEGV crashes this week, on 1.4.3 (422), on M1 running 12.4.

[melonamin]

@bobobox plz give me an example plugin, I can't catch this

[joeshaw]

I run 3, but I can only give you two of them.

utc.5s.bash

#!/bin/bash
exec date -u +':globe: %H:%M | size=13'

stock-ticker.30s. -- This is a Go program. I'm including the source below but if you want me to attach a binary instead I can do that. If you have go installed, drop it in a directory and run go build -o stock-ticker.30s. .

package main

import (
        "encoding/json"
        "fmt"
        "io/ioutil"
        "log"
        "net/http"
)

type result struct {
        Chart struct {
                Result []struct {
                        Meta struct {
                                Symbol    string  `json:"symbol"`
                                Price     float64 `json:"regularMarketPrice"`
                                PrevPrice float64 `json:"chartPreviousClose"`
                        } `json:"meta"`
                } `json:"result"`
        } `json:"chart"`
}

const urlFmt = "https://query1.finance.yahoo.com/v7/finance/chart/%s?interval=1d"

func main() {
        symbols := []string{"AAPL", "^SPX"}
        for i, s := range symbols {
                u := fmt.Sprintf(urlFmt, s)
                resp, err := http.Get(u)
                if err != nil {
                        log.Fatal(err)
                }
                defer resp.Body.Close()

                body, err := ioutil.ReadAll(resp.Body)
                if err != nil {
                        log.Fatal(err)
                }

                if resp.StatusCode != http.StatusOK {
                        log.Fatalf("%d: %s", resp.StatusCode, body)
                }

                var r result
                if err := json.Unmarshal(body, &r); err != nil {
                        log.Fatalf("%v: %s", err, body)
                }

                price := r.Chart.Result[0].Meta.Price
                prev := r.Chart.Result[0].Meta.PrevPrice
                net := price - prev
                pct := net / prev * 100

                var arrow string
                if net > 0 {
                        arrow = "↑"
                } else if net < 0 {
                        arrow = "↓"
                }

                if i == 0 {
                        fmt.Printf("%s $%.2f%s | size=13\n", s, price, arrow)
                        fmt.Println("---")
                }
                fmt.Printf("%s %.2f %+.2f (%+.2f%%)\n", s, price, net, pct)
        }
}

The third is another Go program that runs every 5s. Unfortunately I can't share that one.

[melonamin]

@joeshaw hey, I think I'm ready to give up...

I've been running these two plugins on two different machines(Intel and M1) for two weeks straight, with at least 20 sleep-wake cycle on both of them. It just works.🤷‍♂️

[joeshaw]

@melonamin Is there anything more I can do to help? A debugging build, additional logs, anything like that?

I appreciate your effort trying to reproduce it so far.

[melonamin]

I’m researching how to add a better crash reporting at the moment

[melonamin]

@joeshaw I've added crash reportiing, please try running this build SwiftBar.1.4.4b448.zip

[joeshaw]

Just as an FYI, I've been running this build for about a week now without any crashes. I'll update if and when I see one.

[melonamin]

Hey @adamkdean, since you are expiriencing a simmilar issue in xbar, I would love you to take this build of SwiftBar for a spin.

[adamkdean]

@melonamin once I'm done with testing xbar, I'll give SwiftBar a whirl too and get back to you

[melonamin]

@joeshaw please tell me it's been rock solid for you all this time, I want to release 1.4.4 soon-ish

[joeshaw]

@melonamin No issues here. I feel comfortable with it, and I think you can probably close this issue now. I'll reopen it or file another if I start seeing the issue again. I'm sorry we couldn't precisely identify it!