Rolf Kalbermatter

The change to "librarize" all OpenG functions is a real change in terms of requiring any and every caller to need to be recompiled. This can't be avoided so I'm afraid you will either have to bite the sour apple and make a massive commit or keep using the last OpenG version that was not moved to libraries (which in the long run is of course not a solution).

And what is the program on your ESP32 doing? Does it even listen on the according serial port? Does it know what it should do when seeing an *IDN?<new line> on that port? What does it send back when seeing that command? The ESP32 is a capable microcontroller board but it needs a program that can implement the reading of your sensors and react to commands from your LabVIEW program and send something back. And that program needs to be implemented by you in one of the supported programming languages for the ESP32. Most likely you will want to use ESP-IDF as a plugin in either Eclipse or VSCode.

The only thing I have found to work is to maintain separate projects for 32-bit and 64-bit and have them each build into a seperate location on disk. Anything else is going to mess up your projects almost beyond repair possibilities. That applies to both the projects to build your PPLs, possibly one project per PPL in the case I worked on, as well as for the applications using those PPLs. Using Symlinks to map the build locations of the PPLs to a common path that is referenced by all the projects making use of PPL dependencies (including PPL build projects), helps with maintenance of the projects as they all only need to reference a general location for dependencies rather than an architecture specific location.

5.0.1 and in the meantime 5.0.2 has been since released. One issue, but that is not really new and existed before: Don't disable mass compile after install, it may take some time but it sure fixes stale shared library paths in the VI and I have so far not found a way that makes those paths automatically fixup at package creation, since the path seems to need to be absolute. The two possible approaches I'm currently considering: 1) use a so called Symbolic Path (/<user.lib>/_OpenG.lib/lvzip/lvzip.*). Disadvantage: - only works if installed into default location 2) use Specify Library Name on diagram for the Call Library Node and calculate its path at runtime. Disadvantage: - makes the shared library be not a visible component to the VIs, so that the shared library needs to be added explicitly in every application/shared library/assembly/source distribution build in order to be available in such - extra execution time for the dynamic calculation of the path

If you use the chroot trick that NI/Digilent did for the Linx/Hobbyist Toolkit it is theoretically doable but far from easy peasy. And you still have to sit down with the NI/Emerson lawyers as I told you before. However I doubt you want to run your Haibal library in a specially compiled Debian kernel running in a chroot inside your Jetson Linux distro. That is necessary since the entire NI library tree and LabVIEW Runtime is compiled for the arm softeabi that NI uses for their Xilinx XC7000 based realtime targets. And yes you can NOT run a LabVIEW VI without LabVIEW runtime on the target! Period! And that NI did put the NI Linux RT source online has nothing to do with that they want to be nice to you or let you build your own LabVIEW realtime hardware but simply because it is the most easy way to comply with the GPL license requirements for the Linux kernel. But those license requirements do not extend to any software they create to run on that Linux kernel, since the kernel license has an explicit exemption for that. Without that exemption there would simply not be any commercial software to run on Linux. And I understand what you want but that does not make it feasible. I WANT to win the jackpot in the lottery too but so far it never happened and quite certainly never will. 😀

If and how the DLL uses exceptions is completely up to the DLL and there are actually several way this could work but only the DLL writer can tell you (if it is not explained in the documentation). Windows has a low level exception handling mechanisme that can be used from the Win32 API. It is however not very likely that a DLL would make use of that. Then you have the structured exception handling or its namesakes from different C++ compilers. And here things get interesting since each compiler builder was traditionally very protective about their own implementation and were very trigger happy about suing anyone trying to infringe on the related patents. It means that GCC for a very long time could not use the Microsoft SEH mechanism and therefore developed their own method that was avoiding Microsoft patents. So if your DLL uses exceptions, and doesn't handle them before returning from a function call to the caller, you might be actually in a bit of a bind as you really need to know what sort of exception handling was used. And if you use a different compiler than what was used for the DLL, you will most likely be unable to properly catch those exceptions anyhow, causing even more problems. Basically, a DLL interface is either C++ oriented and then needs to be interfaced by the same compiler that was used for the DLL itself anyhow, or it is a standard C interface and then it should NEVER pass unhandled exceptions to the caller since that caller has potentially no way to properly catch them. One exception are Win32 exceptions that the LabVIEW Call Library Node actually is prepared to catch and turn into the well feared 1097 error everybody likes so much, unless you disable the error handling level completely in the Call Library Node configuration dialog. 😁 Your example code, while probably condensed and not the whole thing, does however ignore very basic error handling that comes into play long before you even get into potential exceptions. There is no check for the Load_Extern_DLL() to return a valid HANDLE. Neither do you check for the function pointers you get from that DLL to be all valid. p2_CallbackFunction_t is rather superfluous and a bit misleading. The _t ending indicates it to be a type definition but in fact you simply declare a stack variable and assign the reference to the CallBack function to it. Luckily you then pass the contents of that variable to the function, so the fact that that storage is on the stack and will disappear from memory as soon as your function terminates is of no further consequence. But you could just as well simply pass the CallBack function itself to that function and completely forget about the p2_CallbackFunction_t variable declaration. Once your function returns, you have indeed no way to detect exceptions from the DLL anymore as there is no stack or call chain on which such an exception could be passed up. The way this should be done is by the DLL handling all exceptions internally and passing an according error indication through the CallBack function in an error variable. It can't use the CallBack function to pass up exceptions either since the CallBack function is called by the DLL, so the exception handling can't go from the callback to LabVIEW but only from the callback to the caller, which is indeed yes ... big drumrolls ... the actual DLL. If your DLL doesn't catch all exceptions properly and handles them by translating them to an error code of some sort and passing that through the callback or some other means to the calling application, then it is not finished in terms of asynchronous operation through callbacks. Exceptions only can pass up through the call chain, but if there is no call chain such as with a callback function, there is no path to pass up exceptions either.

Hmm, not trying to criticize you but having 100 (or even 25) little windows that all display data and allow control too, seems to me to be a pretty difficult UX. It's definitely not something I would immediately turn to. Probably would have more like a list box that shows the information for each device, possibly in a tree structure, and letting the user select one and then make the controls for that available in a separate section of the screen where the control will be specific to the selected device. Shaun's example, while nice technically, shows the difficulty of that approach very well even without much of user control. The graph in there is pretty much way to small for any usable feedback.

I would likely use a Table or MultiColumn List Control.

The whole ADS library overhead in an application adds about 0.0something seconds to the whole build time of any project. As long as you have a linear hierarchy in object class dependencies, there is virtually no overhead at all beyond what you would have for the involved number of VIs anyhow. Once you happen to create circular dependencies the app builder will go into making overtime to resolve everything properly and your build times start to grow into the sky. At some point you can get all kinds of weird build errors that are sometimes almost impossible to understand. Untangling class hierarchies is in that case a very good (and usually also extremely time intensive) refactoring step to get everything to build nicely again.

For this type of functionality it is absolutely not Booh. 😀 It is OOP in the sense that it uses LabVIEW classes as an easy means of instantiating different driver implementations at runtime. One interface (or a thin interface class in pre LabVIEW 2020) and then a child implementation to call the DLL, one to call VISA USB Raw and possibly others. On top of that one driver as a class to do MPSSE I2C and another for MPSSE SPI. That way it is very easy to plugin a different low level driver depending what interface you want to use. D2XX DLL interface on Windows, D2XX so interface or D2XX VISA USB Raw on Linux and LabVIEW RT. With a little extra effort in the Base Class implementation for a proper factory pattern implementation, the choice which actual driver to load can be easily done at runtime. I did the same with other interfaces such as Beckhoff ADS, one driver using LabVIEW native TCP, another interfacing the Beckhoff ADS DLL, one for Beckhoff ADS .Net and one for Beckhoff ADS ActiveX, although the ActiveX one has a bug in the type library that tells LabVIEW to use a single reference byte as data buffer for the read function and there is no way to overwrite the type lib information except patching the ActiveX DLL itself. The Typelib should declare that as an array datatype but instead simply declares it as a byte passed by reference. The same thing for a C compiler but two very different things for a high level type description. The base class implements the runtime instantiation of the desired driver and the common higher level functionality to enumerate resources and read and write data IO elements using the low level driver implementation. For platforms that don't support a specific "plugin" you simply don't install that class.

USB Raw will of course require your LabVIEW code to use the right USB control endpoints, pipes and binary bit patterns and if the new chip is not 100% upwards compatible with the old, that means a different driver (or one with conditional code based on the chip as detected at communication initialization). Without a detailed specification of the low level USB protocol for the new chip that is not gonna work. Using the FTDI provided D2XX driver should fix that as that driver somehow makes sure to talk to each chip as needed, as long as FTDI supports that chip with this driver. Some of the FTDI chips need a different driver such as D3XX. It's definitely easier than trying to do USB Raw communication but requires a bit of C programming knowledge to be able to figure out the proper Call Library Node configuration for each function. The existing LabVIEW wrapper as provided from FTDI is outdated and only supports 32-bit, possibly even with a bug or two even in 32-bit, but it is a starting point.With a little love it can definitely be made to work for modern LabVIEW in 32-bit and 64-bit and for Windows and Linux. I may revisit my earlier attempts of writing a fully LabVIEW OOP based solution for this with pluggable low level driver using USB Raw or FTDI D2XX and a pluggable high level driver for the I2C and SPI MPSSEE modes of the chips that have such an engine build in, but I haven't currently a use case for this and that tends to push this project always towards the end of the "projects that are nice to develop" queue.

It may be possible, but a few 100 subpanels somehow makes my alarm nerve tingle quite strongly! Sub Panels are not the light weight thing like Windows windows, where they made each control in a classic Win32 application its own sub-window. And they moved away from that idea with more modern UI frameworks too, and likely for some reason. You really may be stressing LabVIEW's window management capabilities beyond reasonable borders with so many subpanels present at the same time.

Not just for Linux. In Windows it is the same. But I'm not sure, I never used the D2XX driver on Linux until now. My understanding is that as long as you do not open a VCP interface for a device, the hardware is not really in use and should be accessible by the D2XX driver. In the worst case you may have to configure udev or whatever is used on your system to manage automatic mounting of USB devices, to not install a particular device as tty device, for instance based on the serial number or similar. By default most Linux systems are nowadays configured to automatically install any device they have a standard driver for as the according device. In the past this was often the opposite where you had to mount the device manually or add an according rule to udev for automatic mounting at boot time.

That's not how those drivers work. The device drivers for the FTDI chips (the RS-232422/485 at least) installs in Linux as a serial port driver. That means that they get installed as serial ports under /dev, usually as /dev/ttyUSB* with increasing number and can be accessed through NI-VISA just as any other serial port. There is no exported API in these drivers that you can interface to with a Call Library Node. If you want to access the MPSSE engine in those chips (to do bit bang, I2C or SPI for instance) instead of the Virtual Comm Port interface the standard Linux driver provides, you'll have to hunt down the D2XX driver for your Linux distribution and install that instead. https://www.ftdichip.com/old2020/Support/Documents/AppNotes/AN_220_FTDI_Drivers_Installation_Guide_for_Linux.pdf This document is from 2017. Your Linux distribution may have changed things considerably in the meantime, so your mileage may vary, but the general distinction between standard VCP and FTDI proprietary interfaces to the chip through the D2XX driver remains. There exists LabVIEW libraries to interface to the D2XX shared library and also to access the two FTDI provided wrapper libraries to do MPSSE SPI and MPSSE I2C but they are heavily Windows minded, meaning that they try to link to the .DLL version instead of the .so and there might be other API specific details that differ between Windows and Linux. Also last time I checked the two MPSSE VI interfaces here on LAVA were not cleanly prepared to handle 64-bit shared libraries as they were developed when 64-bit LabVIEW was still in its early days and nobody bothered to install and use that.

I would actually suggest to implement it properly by extending the File Dialog Node. A double NULL terminated string array that is simply passed to a Windows API function is hardly a safe and proper LabVIEW datatype! 😁 And of course you saw it already coming, add support for that on Linux too 😁