Prerequisites

• [x] I am running esp32 board version 1.0.4
• [x] I checked the documentation and found no answer
• [x] I checked to make sure that this issue has not already been filed
• [x] I checked to make sure I'm running the latest build

Expected Behavior

Upon tapping on "PoS Options" in the captive portal menu, I want to be able to see input boxes where I can fill in info on my wallet id, lnurlPOS string, etc.

Current Behavior

Upon tapping on "PoS Options", after a couple of seconds I get a response saying "authentication failed".

If I attempt to access it from my browser and not the captive portal, I am asked to enter a username and password upon tapping on the "PoS Options" option in the menu.

Failure Information (for bugs)

Gathering from this thread iOS (and Android 10 onwards) considers this AutoConnect request to be insecure (because HTTP) and thus disallows it.

The following pages might help with finding a solution:

• Unauthorize error without prompting the login dialog.
• Turn on the debug log options
• AutoConnectConfig API

Steps to Reproduce

1. Launch the AP from ESP32 board by pressing * within 2 seconds of starting
2. On an iPhone, tap on Settings > Wi-Fi > PoS-xxxxxxxx (under Other Networks)
3. Enter password ToTheMoon1 (subject to change based on build)
4. Tap on the hamburger menu on top right
5. Tap on PoS Options

You can also access the portal on a browser after cancelling out of the captive portal by selecting "Use Without Internet" and then navigating to http://172.217.28.1 It may ask you for a username and password if accessed via a browser.

Context

This issue is only relevant for those attempting to access the portal on any Apple devices.

Failure Logs

If a POS URL contains a : followed by a port number, the QR code can't be scanned. I tried in my mobile Lightning wallets and also with a QR app on my phone. If I delete the : und the port, the QR ist scannable (but wrong).

When I add an amount of sats in the POS, and read the QR Code in my wallet, the amount is different. I have tested it with two different wallets: LN Bits and BlueWallet, and the value was different from the POS in both of them

For example, when I inserted 1sats, my wallet read as 7sats

Issue

When having 2 menuItems enabled, there was no way to get to the second item. At first it seamed to be a bug with the keypad, but since the '*' was working correctly on other screens it didn't make sense.

Bug

After reviewing the code, I've noticed that the menuLoop function was resetting the loop in a way that didn't cover well all the possibilities. Inside the "btnloop" while loop (Line ~1258) you will find that the menuItemNo is iterated with the complete possible menu in mind [0-3] while the resetting condition (Line ~1274) was compared to the menuItemCount which contemplated only active menuItems. This doesn't comply when you have only LNUrlPOS and LNUrlATM items, since the menuItemCount is 2, but LNUrlATM is 3.

Solution

largestMenuItemChecked is a new variable that holds the largest item that is active in the menu. When loading the items checked we update this variable and then use it to compare with in the condition to reset the menuloop.

Variable is started at 0 so there is no need to add any code for menuItem 0, nor validate anything if this item is deactivated later due to lack of wifi connection.

There is already another pull request that solves the issue with an alternative solution. I'm proposing this one since it might be better aligned with the current design. If the other gets merged I will delete this one. I appreciate any comments. Awesome project!!

There are some combinations of enabled menu items which stop the selection with '*' key from working.

As an example, where the menu items enabled are:

• LNURLPoS
• LNURLATM

In this case, only LNURLPoS can be selected and pressing '*' does not advance the selection.

Where LNURLPos & LNURLATM are the only menu items available, it's not currently possible to select LNURLATM. An update to the logic for selecting the next item on a '*' key press fixes this.

There are two compiler warnings with regard to NULL values which are fixed here by introducing a sentinel value for USB_POWER.

The keypad delay is parameterised and reduced to 50ms. This makes the keypad feel more responsive - no keybounce issues have been observed.

This has been tested with onchain and lnurlpos but not lnpos as I am getting server errors when the device is checking for payment on the invoice.

Also, not implemented for LNURLATM yet as I don't know what that even is :p

Fixes the issue that currently the mempool.space link is given in the QR as

https:/mempool.space/address/bc1...

Changed to https:// in all 3 .ino-s.

(Previously could give the custom link as /mempool.space to work around this)

Larger hole to allow for variation in screen placement on ttgo board 0.2mm gap between case upper and lower lips to accomodate less precission on some 3d printers

Some thgings I came acrross while setting up my barebones LNPoS:

• Invoice memory length was incorrect and led to a NoMemory when fetching the JSON
• LNURRL-Withdraw QR code was too small, applied the same fix as in #26
• Improved and adjusted the display of the screens

In the lnPoSTdisplay version, the payment success screen shows the message PAYED. The correct English term would be PAID.

Change tested with TTGO device:

Would like to have an option to customize the "checkout page" when completing the transaction. For now, the pin is showing at LNpos device and the link for the one that did the payment. Only the pin is showing when clicking the link. This page could have some more details:

• Summary of transaction
• Timestamp
• Sats paid
• Fees paid in sats
• Sats value in local fiat currency
• maybe some information of the company / client who received the sats
• an Email-Form to send all this information to the person who paid and as cc or bcc to the company.

For me personally living in Germany it would be awesome to always get all transaction details from above as email.

All this is available on BTC-PayServer. Getting the transaction details should not be a problem, I think. Creating a form is simple HTML.

For me running Umbrel I was not able to find a path where I can customize the page showing when transaction is done.

How the unification was done:

mkdir -p lnPoS/libraries/

cp lnPoSBareBones/lnPoSBareBones.ino lnPoS/lnPoSBareBones.cpp.h cp lnPoSM5Stack/lnPoSM5Stack.ino lnPos/lnPoSM5Stack.cpp.h cp lnPoSTdisplay/lnPoSTdisplay.ino lnPoS/lnPoSTdisplay.cpp.h cp -R lnPoSM5Stack/libraries/* lnPoS/libraries/

• built lnPoS/lnPoS.h based on headers from lnPoSBareBones/lnPoSBareBones.ino, lnPoSM5Stack/lnPoSM5Stack.ino and lnPoSTdisplay/lnPoSTdisplay.ino
• built lnPoS/lnPoS.ino entry point
• built lnPoS/hardware_device.h that allows configuration of the hardware device
• reworked User_Setup_Select.h of the TFT_eSPI library to get the hardware device from hardware_device.h and then include the appropriate file from User_Setups/
• cp lnPoSBareBones/libraries/TFT_eSPI/User_Setup.h lnPoS/libraries/TFT_eSPI/User_Setups/Setup0_lnPoSBareBones.h

Next steps:

• test the code on BareBones, M5Stack and Tdisplay to make sure nothing broke
• gradually reduce code duplication by moving identical functions to lnPoS.ino
• merge the remaining differing but similar functions so they can also be moved to lnPoS.ino

Regarding the unusual .cpp.h extension

The original .ino files have been moved to .cpp.h files, which is an unusual file extension.

The reason is:

• a file extension that the Arduino IDE will open by default when opening the project, so we need either .h, .cpp or .ino
• the extensions .cpp and .ino can't be used because these files will always be included in the compilation so we can't use a selective #include for them, resulting in "duplicate definition"s
• so only .h remains
• to make it clear that these are not regular header files (without code) but actual code files, the extension is .cpp.h

Comparing old with new code:

Also see attachments.

This allows you to verify that the code changes were really minimal:

cd ~/Arduino/LNPos/ # go to the top-level

# Inspect the code changes:

diff -Nabur deprecated_since_unification/lnPoSM5Stack.ino lnPoS/lnPoSM5Stack.cpp.h | tee /tmp/unification_lnPoSM5Stack.ino.code_diff
diff -Nabur deprecated_since_unification/lnPoSTdisplay.ino lnPoS/lnPoSTdisplay.cpp.h | tee /tmp/unification_lnPoSTdisplay.ino.code_diff
diff -Nabur deprecated_since_unification/lnPoSBareBones.ino lnPoS/lnPoSBareBones.cpp.h | tee /tmp/unification_lnPoSBareBones.ino.code_diff
less /tmp/unification_*code_diff

# Inspect the header changes:
diff -Nabur deprecated_since_unification/lnPoSM5Stack.ino lnPoS/lnPoS.h | tee /tmp/unification_lnPoSM5Stack.ino.header_diff
diff -Nabur deprecated_since_unification/lnPoSTdisplay.ino lnPoS/lnPoS.h | tee /tmp/unification_lnPoSTdisplay.ino.header_diff
diff -Nabur deprecated_since_unification/lnPoSBareBones.ino lnPoS/lnPoS.h | tee /tmp/unification_lnPoSBareBones.ino.header_diff
less /tmp/unification_*header_diff

unification_lnPoSBareBones.ino.header_diff.txt unification_lnPoSTdisplay.ino.header_diff.txt unification_lnPoSM5Stack.ino.header_diff.txt unification_lnPoSBareBones.ino.code_diff.txt unification_lnPoSTdisplay.ino.code_diff.txt unification_lnPoSM5Stack.ino.code_diff.txt

Once set up, one cannot remove the onchain option as the key file will still be present.

It should only be available in the setting contains an actual xpub, so that the onchain option isn't available if the setting gets cleared out.

Alternatively the key file should be deleted when the setting is cleared out.

I tested this project on a TDisplay and everything works fine except the LNPoS Service. (LNURLPoS and LNATM works fine) My WiFi is set up correctly.

I debugged it with a connected serial console:

connecting to LNbits server legend.lnbits.com 
10:12:47.240 -> failed to connect to LNbits server legend.lnbits.com 
10:12:50.264 -> menuLoop

I don't know why this happens (yet).