Watching program flow in a time-sensitive situation

I’m working on some unit test in Java (well, Kotlin) for an Android project. When I set a breakpoint, the failure mode changes because there are timeouts and races going on. This is not an ideal situation, but I’m debugging someone else’s code here. So I found a clever way to log what code is getting called, and from where, by using breakpoint actions. Just put `new RuntimeException().printStackTrace()` in as a breakpoint action! Now your log shows when code got hit, and why it got there, but doesn’t actually stop execution, so it does less intrusive shift to the timeouts and race-conditions.

Receipt Validation on iOS

This has been a pretty major headache: Trying to validate a receipt for an in-app purchase on iOS.


  1. Use cocoapods OpenSSL-Universal pod to get openssl
  2. Copy Apple’s suggested code, but add the key missing call OpenSSL_add_all_digests() prior to PKCS7_verify()
  3. This last thing is what fixes the mysterious error that reads, “error:2107D06D:PKCS7 routines:PKCS7_BIO_ADD_DIGEST:unknown digest type”

Now, to explore the contents of my ASN1 package!

Kotlin, please allow return from closures

Edit: After writing this, I was helped by Andrey on Twitter and now I know about the return@

The simple answer is:

sequence.flatMap {
if (it.containsKey("connnected") {
return@flatMap Observable.just(it.getBoolean("connected"))
return@flatMap Observable.empty()

Here is the original post.

Today’s kotlin frustration… trying to map a Bundle into types. E.g. if the bundle has the key “connected” get it out as a boolean. otherwise, skip the event.

so I do

sequence.flatMap {
if (it.containsKey("connnected") {
return Observable.just(it.getBoolean("connected"))
return Observable.empty()

Except, oh Kotlin, you back-stabbing-friend… closures don’t allow `return`

So I edit it…

sequence.flatMap {
if (it.containsKey("connnected") {

That’s probably obviously wrong to you now, but it took me a long time to see the problem.

And the fix.

sequence.flatMap {
if (it.containsKey("connnected") {
return Observable.just(it.getBoolean("connected"))
} else {
return Observable.empty()

I don’t see why Kotlin disallows the return — and it makes this kind of error much less likely. The fact that closures evaluate to the last expression is handy for doing tiny bits of work, but I find that reactive networks in RxJava are much easier to express with blocks — and if they get more than a few lines long, you sometimes really want the ability to do an early return without introducing the nested scope of an `if` statement.

After installing openvpn I cannot connect to my raspberrypi

So I installed openvpn on my Raspian distro on my Raspberry Pi 2. As soon as I started it up, I lost all connectivity to my Pi.

And I don’t have a display, keyboard, or mouse — all my management has been over the network.

How to recover?

I tried a bunch of things, but here’s what worked, in brief.

  • Power off PI and put SD card in my Mac.
  • Copy the SD card to an image file using `dd`
    • dd bs=32768 if=/dev/rdisk2 of=raspberrypi.raw
    • VBoxManage convertdd raspberrypi.raw raspberrypi.vdi --format VDI

Use Disk Utility to see which device is your card reader. The Pi’s SD card has two partitions, and the first can be mounted by your mac. Note by using ‘rdisk2’ instead of ‘disk2’ it does unbuffered reading which is faster, or so I’m told.

Now we edit the image. If you have a way to open ext4 filesystems already, just remove the files (see rm step below). Here’s how I did it:

  • Install VirtualBox with the latest Finnix Recovery Distro.
  • Add the .vdi file created above to the Finnix VM.
  • Boot Finnix
  • Mount the image
  • mkdir -p /mnt && mount -t ext4 /dev/sda2 /mnt
  • The reason RPi doesn’t boot with open vpn is because ifplugd tries to HOT plug it.  To fix this, explicitly list the real network interfaces in `/etc/default/ifplugd`:
    • Replace
    • INTERFACES="auto"
    • with
    • HOTPLUG_INTERFACES="eth0 wlan0"

    Failling that, you can always just remove all the init scripts that start openvpn.

  • rm /mnt/etc/rc*/*openvpn
  • Unmount
    • umount /mnt
  • Power off Finnix
    • shutdown -h now
  • Convert back to RAW format
    VBoxManage clonehd raspberrypi.vdi raspberrypi.raw --format RAW
  • Copy the image back onto the SD card:
    • sudo dd bs=32768 if=raspberrypi.raw of=/dev/rdisk2

Put the SD card in your Raspberry Pi and reboot!


I’m so excited about Linea, new sketching app for the iPad from Iconfactory. Already we’re seeing some amazing drawings and sketches from users! Creating new software and releasing it into the wild is exciting and thrilling. We’re already hard at work at a dot release, so please give it a try and let us know how you like it.

Now running nginx and secure

So I think I’ve finished migrating my blog to the new server. It’s on Ubuntu 16.04 now, instead of 12, and running on PHP7 instead of PHP5, and on mariadb instead of mysql, and on Nginx instead of Apache.

Lots of little problems came up.

For one thing, the old database was configured so that the character set was latin1 instead of utf8. But wordpress was putting utf8 bytes into it. Which worked fine until I went to migrate the data. Because now-a-days mysqldump delivers data in utf8 encoding by default, so in order to get the bytes out unmolested, I had to specify to dump as latin1. Then, it’s necessary to edit the dumped .sql file and change the latin1 in there to utf8. Quite a headscratcher. If you see accented characters showing up as TWO weird characters, this is probably what happened to you.

ssh user@hostname.tld 'echo XXXXXpasswordXXXXX | mysqldump --default-character-set=latin1 -u root -p database' > data.sql

Next, a shout out to Lets Encrypt which is providing the SSL certificates for my domains. Configuration was incredibly easy.

I’m still running on Linode, which is working reliably — as it has done for the past many years! They no longer let me create an instance of the size I used to use, so my new instance is slightly smaller… when all is said and done I might save a little money!

I’m also trying an experiment where I track changes to my Linux configuration in `/etc` using GIT. Unfortunately, lots of files change in `/etc` for reasons that aren’t meaningful configuration changes… so it will probably be of dubious value.

Now on to the rest of my domains…


Have you tried out the new iOS Messages Stickers yet? I bet you’ve found some that are your favorites: silly emoji or some of your favorite Disney characters? My favorite pack so far is FeelieFaces ( by Sydikos! It’s over two dozen animated faces that show the whole range of human emotion. There are sad face FeelieFaces and happy faces, too! The best one, in my opinion, is the Pirate face who says “Arrrggh!”:


SevSeg Arduino and Easy Driver Stepper Motor

This is a continuation of my previous post. When we left off, I was trying to figure out which pins on the Sparkfun SevSeg display I could repurpose for my own purposes as digital out pins.

I did some research on the ATMega chip, and it is pretty clear that most of the pins on the top row of the SevSeg could be used for other purposes. Having identified which (Arduino) pins were connected to the display, I was able to figure out which pins were connected to the top row of solder points by using a simple programs that displayed a number on the LCD while simultaneously toggling that pins state to high and low. I had some trial and error — for example my multimeter takes about 4 seconds to register a change in voltage and my initial script was changing the voltages faster than that… at first it seemed like nothing was happening! But finally I figured out the pin numbers for the solder points. Only after I did all that did I look in the header file provided by the Sparkfun “board” for the SevSeg and see that they’d provided convenient macros for those points! All in all I found 6 available digital I/O pins.

I decided to connect five pins to the Easy Driver — Enable, Step, Dir, MS1 and MS2. I left MS3 disconnected. This will allow me to control the direction of rotation, turn off power to the motor, and use half, quarter and one-eight stepping modes for slow speeds.

That left one pin for a button. I verified that I could make inputs work by simply connecting one output pin to one input pin by sliding a resistor between the two solder points. When I programmatically set one of them to output HIGH while putting the other into INPUT_PULLUP, I verified that I could read a state change.

Next I soldered connections between the buttons and the EasyDriver and the SevSeg. I’m also using the EasyDriver as a power supply for the SevSeg.

I found some example code for how to drive the stepper motor by sending alternating HIGH and LOW pulses to the STEP pin, each with some delay. This worked perfectly, on the first try. However, when I tried adding the digital display to show how many rotations are left, I ran into a problem. The SevSeg.cpp uses delays internally to rapidly illuminate each digit in succession. I also needed carefully timed delays between each STEP I send to the stepper. Arduino doesn’t have any threading support. I was facing having to build some kind of scheduler to allow these two tasks to do small bits of work offset by time delays. Ugh.

But then I discovered the clever pt.h “threads” library by Adam Dunkels. This imposes some pretty strict limitations on your “C” code — e.g. locals are not persistent across WAITs. But the solution to that was the (normally frowned upon) practice of using static/globals for everything! The original SevSeg library was built around a C++ object, which I had to refactor to be a static “C” function — easily done once I made all its member fields public (another no-no, normally!).

The end result: I can control both the motor and the display simultaneously without one (significantly) interfering with the other.

And here is a demo of the finished project: