Last updated on April 10, 2022
open62541 setup using the source code on Linux and build the executable of a minimal open62541 server
This Post is part of the OPC UA Information Model Tutorial.
Contents
open62541 Documentation
This setup guide is based on the official documentation of open62541. It augments and comments the official documentation whenever necessary: https://open62541.org/doc/current/
A printable version of the online documentation can be downloaded from https://open62541.org/doc/open62541-current.pdf for offline use.
Get Source Code
To get the source code and install open62541 from source, use the following steps:
git clone https://github.com/open62541/open62541.git ~/open62541 cd ~/open62541 git submodule update --init --recursive
Build the library
This section refers to https://open62541.org/doc/current/building.html#building-with-cmake-on-ubuntu-or-debian from the official documentation.
You first need to install all required dependencies and then use cmake or ccmake to configure the build environment:
sudo apt-get install git build-essential gcc pkg-config cmake python # enable additional features sudo apt-get install cmake-curses-gui # for the ccmake graphical interface sudo apt-get install libmbedtls-dev # for encryption support sudo apt-get install check libsubunit-dev # for unit tests sudo apt-get install python-sphinx graphviz # for documentation generation sudo apt-get install python-sphinx-rtd-theme # documentation style mkdir ~/open62541/build && cd ~/open62541/build ccmake ..
Note that the following steps use /home/user/install (or /home/ubuntu/install) as the installation path. Make sure that you change it to your username. You can also use the following value instead: $HOME/install.
In the newly opened window press c and then configure the following values. When you are done, press c followed by g :
- configure a custom install path to
/home/user/install- avoid polluting Debian’s system paths with custom libraries
- Set the build type to
RelWithDebInfo- This will create a library which is optimized for Release, but still includes debugging info for additional error investigations
- Enable the Full Namespace Zero
- To build your own information models based on the official OPC UA Specification you most probably need the complete Namespace Zero
Note: ccmake configures cmake which in turn configures make. Consequently, setting options in ccmake is equivalent to setting build flags in cmake.
For more information, visit https://cmake.org/runningcmake/
Example: cmake -DCMAKE_INSTALL_PREFIX=/home/user/install -DCMAKE_BUILD_TYPE=RelWithDebInfo -DUA_NAMESPACE_ZERO=Full ..
Build the library. This will take a few minutes. Transfer the build artefacts to the custom install folder ~/install by calling make install. This is required for building your own custom server based on the installed sources.
cd ~/open62541/build make -j make install
Building the server executable using gcc
This section refers to https://open62541.org/doc/current/tutorial_server_firststeps.html
Create a new directory for your own code ~/myServer. Then copy the tutorial code to your own file:
mkdir ~/myServer cp ~/open62541/examples/tutorial_server_firststeps.c ~/myServer/myServer.c
Now we can build this custom server code by adding the custom installation directory to the gcc command, and by adding additional required libraries
- adding our custom installation paths as include path
-I$HOME/install/include - adding our custom library paths
-L$HOME/install/lib - adding the open62541 library
-lopen62541 - adding
-lmbedtls -lmbedx509 -lmbedcryptofor crypto
cd ~/myServer gcc -std=c99 -I$HOME/install/include -L$HOME/install/lib myServer.c -lopen62541 -lmbedtls -lmbedx509 -lmbedcrypto -o myServer
If you’re building the FULL namespace 0 (as recommended by this guide) be prepared to see the following output
cd ~/myServer
gcc -std=c99 -I$HOME/install/include -L$HOME/install/lib myServer.c -lopen62541 -lmbedtls -lmbedx509 -lmbedcrypto -o myServer
/home/ubuntu/open62541/build/src_generated/open62541/namespace0_generated.c: In function ‘namespace0_generated’:
/home/ubuntu/open62541/build/src_generated/open62541/namespace0_generated.c:117159:15: note: variable tracking size limit exceeded with ‘-fvar-tracking-assignments’, retrying without
117159 | UA_StatusCode namespace0_generated(UA_Server *server) {
| ^
This is just a note that your server was built without some debugging information. As long as you do not intend to debug open62541 itself, this note has no relevance for you. The server executable myServer was build successfully and is ready to run!
For production use, build the server in RELEASE mode.
Detailed explanation of the custom build command (optional)
You can skip this section if you don’t care why we deviate from the official documentation with above build command.
The official documentation is using the amalgamated build. This is still outdated as it is recommended to use the non-amalgamated version. If you follow the official documentation closely, the build command
gcc -std=c99 open62541.c myServer.c -o myServer
will also fail with
gcc -std=c99 open62541.c myServer.c -o myServer
myServer.c:1:10: fatal error: open62541/plugin/log_stdout.h: No such file or directory
#include <open62541/plugin/log_stdout.h>
^~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
compilation terminated.
gcc fails with the first include in myServer.c It seems, like it’s missing the custom install path from its includes. That was to be expected. Verify the assumption and query gcc’s include paths
source: https://stackoverflow.com/questions/4980819/what-are-the-gcc-default-include-directories
# user@debian:~/open62541/build$ gcc -xc -E -v - [...] #include "..." search starts here: #include <...> search starts here: /usr/lib/gcc/x86_64-linux-gnu/8/include /usr/local/include /usr/lib/gcc/x86_64-linux-gnu/8/include-fixed /usr/include/x86_64-linux-gnu /usr/include End of search list.
The custom install path /home/user/install is missing. This is a case of a missing -I (include) option to the gcc command. With -I/home/user/install the directory is added to gcc’s search path. This can be verifyed as follows
# user@debian:~/open62541/build$ gcc -I/home/user/inst -xc -E -v - [...] #include "..." search starts here: #include <...> search starts here: /home/user/inst /usr/lib/gcc/x86_64-linux-gnu/8/include /usr/local/include /usr/lib/gcc/x86_64-linux-gnu/8/include-fixed /usr/include/x86_64-linux-gnu /usr/include End of search list.
Retry
gcc -std=c99 -I/home/user/install/include open62541.c myServer.c -o myServer
Progress! All header files are successfully located, but the linker fails
# user@debian:~/open62541/build$ gcc -std=c99 -I/home/user/inst/include open62541.c myServer.c -o myServer /usr/bin/ld: /tmp/ccgaXZyP.o: in function `mbedtls_pk_get_len': open62541.c:(.text+0x2c37): undefined reference to `mbedtls_pk_get_bitlen' /usr/bin/ld: /tmp/ccgaXZyP.o: in function `reloadCertificates': open62541.c:(.text+0x355d92): undefined reference to `mbedtls_x509_crt_free' /usr/bin/ld: open62541.c:(.text+0x355da5): undefined reference to `mbedtls_x509_crt_init' /usr/bin/ld: open62541.c:(.text+0x355dfb): undefined reference to `mbedtls_x509_crt_parse_path' /usr/bin/ld: open62541.c:(.text+0x355e49): undefined reference to `mbedtls_strerror' /usr/bin/ld: open62541.c:(.text+0x355efb): undefined reference to `mbedtls_x509_crl_free' /usr/bin/ld: open62541.c:(.text+0x355f10): undefined reference to `mbedtls_x509_crl_init' /usr/bin/ld: open62541.c:(.text+0x355f8e): undefined reference to `mbedtls_x509_crl_parse_file' /usr/bin/ld: open62541.c:(.text+0x3560b3): undefined reference to `mbedtls_x509_crt_free' /usr/bin/ld: open62541.c:(.text+0x3560c8): undefined reference to `mbedtls_x509_crt_init' /usr/bin/ld: open62541.c:(.text+0x356123): undefined reference to `mbedtls_x509_crt_parse_path' /usr/bin/ld: /tmp/ccgaXZyP.o: in function `certificateVerification_verify': open62541.c:(.text+0x356282): undefined reference to `mbedtls_x509_crt_init' /usr/bin/ld: open62541.c:(.text+0x3562a9): undefined reference to `mbedtls_x509_crt_parse' /usr/bin/ld: open62541.c:(.text+0x356331): undefined reference to `mbedtls_x509_crt_verify_with_profile' /usr/bin/ld: open62541.c:(.text+0x356425): undefined reference to `mbedtls_x509_crt_verify_with_profile' [...]
Is the mbedTLS library missing? Make sure it’s installed on Debian.
# user@debian:~/open62541/build$ sudo apt-cache search mbedtls [sudo] password for user: libmbedtls-dev - lightweight crypto and SSL/TLS library - development files libmbedtls-doc - lightweight crypto and SSL/TLS library - documentation libmbedtls12 - lightweight crypto and SSL/TLS library - tls library libshadowsocks-libev-dev - lightweight and secure socks5 proxy (development files) libshadowsocks-libev2 - lightweight and secure socks5 proxy (shared library) shadowsocks-libev - lightweight and secure socks5 proxy
Let’s install them
# user@debian:~/open62541/build$ sudo apt install libmbedtls-dev libmbedtls12 Reading package lists... Done Building dependency tree Reading state information... Done libmbedtls-dev is already the newest version (2.16.0-1). libmbedtls12 is already the newest version (2.16.0-1). libmbedtls12 set to manually installed. 0 upgraded, 0 newly installed, 0 to remove and 0 not upgraded.
The output indicates that mbedTLS is already installed. Pass the library files to the linker with the -l flag. Do not confuse it with -L , which is used for folders.
gcc -std=c99 -I/home/user/install/include open62541.c myServer.c -lmbedtls -lmbedx509 -lmbedcrypto -o myServer
Compilation completes successfully.
Execute the test server
# user@debian:~/myServer$ ./myServer
Continue with Step 2 (Windows)
Continue with Step 2 (Linux)
Hi Stephen!
Thank you for documenting open62541! It is a lot of help in understanding the concepts! Great job!
However, I am unable to execute the sample server.
There are variations in custom install path
in ccmake -> /home/user/install
in image -> home/ubuntu/install
in the launch command -> $HOME/install/include
is this on purpose?
Upon executing the sample server, errors are
-> open62541/plugin/log_stdout.h: No such file or directory when following the instructions
-> open62541.h: No such file or directory when trying to include “open62541.h” as header
Kindly request your input!
Hi audupi,
you are correct, the username displayed in the image image is inconsistent with the rest of the documentation. This documentation consistently uses
useras username, while the image usesubuntuas username.The image should show
/home/user/install.Be sure to replace
user(orubuntuin the image) with the actual username you’re using on your Linux computer!Note:
$HOMEis a well-known Linux variable for your current user’s home directory. It’s equivalent to typing~in your terminal. Let us assume that your username on your Linux computer isaudupi. Therefore all following paths will point to the same directory:$HOME/install~/install
/home/audupi/install
If you are not familiar with Linux, I recommend to use
/home/audupi/install. It’s the least error-prone way to describe the path.Hi there!
Can we have two compiled versions of open62541 installed in the same machine?
Version1, following the same procedure as above and building it in /build.
(/home/open62541/build)
Version2, is a ‘single file release’ version which generates open62541.c & open62541.h in /build_sfr.
(/home/open62541/build_sfr)
Will this lead to any issues in finding the correct “open62541 package” in CMake when building new opcua servers ?
Thanks in advance.
Hi Audupi,
you should be able to have them side-by-side. I have not tested that myself though.
In case you run into any CMake issues I suggest you find help at the appropriate places (forums, IRC channels, stackoverflow, …)
Good luck!
PS:
A remark towards your question:
Will this lead to any issues in finding the correct “open62541 package” in CMake when building new opcua servers ?
You will most likely not get an answer to this this hypothetical question from anyone, because
1) it shows that you have not bothered to try it yourself (and that may be considered lazy)
2) you are forcing anyone who wants to help you to make probably incorrect assumptions about your setup thus wasting a lot of time of others
In general, better questions will give you better feedback/help. Try it and then ask about specific error messages (and give sufficient background information).
Hi there,
Very nice tutorial, I like appreciate the detailed explanations and the explicit note why they are interesting.
I assume the build prerequisite are copied from the open62541 description.
Those two commands did not work for me:
sudo apt-get install python-sphinx graphviz # for documentation generation
sudo apt-get install python-sphinx-rtd-theme # documentation style
The python-sphinx documentation states that the package is python3-sphinx now.
Building the Server leads to this message for me:
/home/alex/open62541/build/src_generated/open62541/namespace0_generated.c: In function ‘namespace0_generated’:
/home/alex/open62541/build/src_generated/open62541/namespace0_generated.c:117159:15: note: variable tracking size limit exceeded with ‘-fvar-tracking-assignments’, retrying without
117159 | UA_StatusCode namespace0_generated(UA_Server *server) {
| ^
I am working on a fresh setup Ubuntu VM, so I don’t expect any version issues from that side.
Cheers and merry christmas,
Alex
Hi Alex,
thank you for your valuable feedback!
Regarding apt-get:
==============
I’ll leave the guide as is, because that’s what I tested and confirmed to work. I’ll update this guide with your suggestions when I find time to test it myself. Your input is appreciated!
Regarding variable tracking size limit exceeded:
====================================
This warning is only relevant for you if you want to debug code.
You can increase the variable tracking size limit like so:
https://stackoverflow.com/questions/23499909/adjust-variable-tracking-assignment-length
You can read more about variable tracking size limit here:
https://access.redhat.com/documentation/en-us/red_hat_enterprise_linux/6/html/developer_guide/ch-debug-vta
Above link also shows that it can be disabled with ‘-fno-var-tracking-assignments’
Hey Opc Rocker,
I try to learn OPC UA but I have on my raspberry a problem with the gcc command….
so when I try the steps, without the “Detailed explanation of the custom build command (optional)”
I get this erros:
root@raspberrypi:~/myServer# gcc -std=c99 -I/home/MYUSER/install/include myServer.c -lmbedtls -lmbedx509 -lmbedcrypto -o myServer
/usr/bin/ld: /tmp/cc643rFr.o: in function `UA_ServerConfig_setMinimal’:
myServer.c:(.text+0xb0): undefined reference to `UA_ServerConfig_setMinimalCustomBuffer’
/usr/bin/ld: /tmp/cc643rFr.o: in function `stopHandler’:
myServer.c:(.text+0x138): undefined reference to `UA_Log_Stdout’
/usr/bin/ld: /tmp/cc643rFr.o: in function `main’:
myServer.c:(.text+0x168): undefined reference to `UA_Server_new’
/usr/bin/ld: myServer.c:(.text+0x174): undefined reference to `UA_Server_getConfig’
/usr/bin/ld: myServer.c:(.text+0x18c): undefined reference to `UA_Server_run’
/usr/bin/ld: myServer.c:(.text+0x198): undefined reference to `UA_Server_delete’
collect2: error: ld returned 1 exit status
with details instruction it ends that it don´t find the open62541.h file.
I think but don´t know that I have a problem with the ccmake options.
There is after editing the setup a flag error (after click c on the keyboard)
but the make install command after ccmake works without error
Can you help me with my beginner questions ?
Thank you 🙂
Hi,
I noticed that you’re missing
open62541.cin your command, which seems to be taken from the “detailed” section.Compare your commands with the given instructions carefully.
Good luck!
Hi Benedict,
Thanks for the fast answer 🙂
yes but I don´t find the open62541.c file in my filesystem were can I get this file?
Tomorrow I will try it again, maybe it was only a small mistake on my side…
Hi there,
Today I try it again and again it ends up with an #include error
first try it was again as root
second try as myuser
myuser@raspberrypi:~/opcUaServer $ sudo gcc -std=c99 -Ihome/myuser/install/include -Lhome/myuser/install/lib opcUaServer.c -lopen62541 -lmbedtls -lmbedx509 -lmbedcrypto -o opcUaServer
opcUaServer.c:23:10: fatal error: open62541/plugin/log_stdout.h: No such file or directory
#include
in the inlcude path the files are there and the user has permission to that files…
I try also with the ccmake to build also the examples and start directly the simpleServer. Thats works fine.
Have you any idea what I making wrong?
Thank you
regards Klemens
sorry I´m just a stupid student 😉
sometimes the small things make big problems
I write home/…. but path starts with /home/…
sorry for wasting your time
Hi Klemens,
I am glad you got it up and running. As I said
Compare your commands with the given instructions carefully.😉 No time wasted on my time.Good luck!
Hi Benedict,
I try it again and again, but on my raspberry Pi it ends up with the following error after round about 2hours, with the gcc command. To you have any Idea or workaround? Or do I need an Raspberry with more Ram? I have now 2GB Ram…
gcc -std=c99 -I$HOME/install/include -L$HOME/install/lib myServer.c -lopen62541 -lmbedtls -lmbedx509 -lmbedcrypto -o myServer
/home/user/open62541/build/src_generated/open62541/namespace0_generated.c: In function ‘namespace0_generated’:
/home/user/open62541/build/src_generated/open62541/namespace0_generated.c:117159:15: note: variable tracking size limit exceeded with -fvar-tracking-assignments, retrying without
UA_StatusCode namespace0_generated(UA_Server *server) {
^
gcc: fatal error: Killed signal terminated program lto1
compilation terminated.
lto-wrapper: fatal error: gcc returned 1 exit status
compilation terminated.
/usr/bin/ld: error: lto-wrapper failed
collect2: error: ld returned 1 exit status
Hi,
1) Start developing on a desktop or laptop computer instead of Raspberri Pi. That way, mistakes will not cost you as much time. If you’re using Raspian, use Debian for your desktop/laptop for simpler transition later on.
2) Regarding the message “note: variable tracking size limit exceeded with -fvar-tracking-assignments” That is just a note that some debugging capabilities will be limited [1] Nothing of concern for you right now.
3) What really makes your build process fail is the fatal error “gcc: fatal error: Killed signal terminated program lto1” This is related to the Link Time Optimization (LTO) step during build.[5] I never had this issue myself but it seems you’re not alone [3][4][6]. I share your suspicion: The linker is running out of memory (OOM) and gets killed by Linux (Killed signal terminated program lto1 is a strong indicator for that). You can try to disable LTO with -fno-lto [5] or try to limit the number of parallel linking jobs to one with -flto=1 [5] (Note: this is related to “MAKE -j”) if that isn’t already the case. If you can get your hands on a Raspi with more RAM, try that too!
Let us know of your results & solutions.
Good luck!
[1] https://access.redhat.com/documentation/en-us/red_hat_enterprise_linux/6/html/developer_guide/ch-debug-vta
[2] Your build process is at its final steps. Maybe you can push through the build process on Raspberri Pi for “quick” results, but for further development I strongly recommend 1)
[3] See comment at: https://opcua.rocks/step-9-using-a-custom-uanodeset-with-open62541/
[4] https://github.com/open62541/open62541/issues/3267 <- This seems to refer to the documentation on building open62541 on Raspberri Pi at https://www.kalycito.com/running-opcua-on-raspberry-pi/
[5] https://man7.org/linux/man-pages/man1/gcc.1.html <- Search for "-flto" [6] https://www.mail-archive.com/meta-arago@arago-project.org/msg11260.html
Hi Benedict,
Thanks for your answer,
Here some Updates
Before I start with your example, I build the firststep server like kalycite[4] and thats works. After I read this source in your answer I´m wondering how I do this and check the ccmake settings…
I change your example without the change from namespace_zero to full.
Now I can build the server on the raspberry
after that I try to build the server on my mac and on a vm with debian
both ends with this error:
gcc -std=c99 -I$HOME/install/include -L$HOME/install/lib myServer.c -lopen62541 -lmbedtls -lmbedx509 -lmbedcrypto -o myServer
/home/parallels/open62541/build/src_generated/open62541/namespace0_generated.c: In function ‘namespace0_generated’:
/home/parallels/open62541/build/src_generated/open62541/namespace0_generated.c:117159:15: note: variable tracking size limit exceeded with -fvar-tracking-assignments, retrying without
UA_StatusCode namespace0_generated(UA_Server *server) {
but it works also if I don´t change the namespace-zero settings…
Whats kind of problems came in, when I work without the full namespace_zero?
Maybe you need to know what I need?!:
My Goal is to build some opc ua servers for a fischertechnik demonstrator. The servers should work together and then I have an edge device with an opc ua client / server for the HMI and Communication outside the local net.
Hi,
1) You need to understand what NAMESPACE_ZERO does [1]. Namespace 0 is anything in namespace “http://opcfoundation.org/UA/” [3] That’s where the OPC Foundation defines standard server nodes and all basic node types you may want to use in your model. If you follow the instructions from Kalycito [2] NAMESPACE_ZERO is REDUCED by default. Depending on your needs, this may be sufficient. You may move ahead with REDUCED (although there is no need too; see my second point in this comment), but be prepared to revisit that decision later on.
2) Again, the message “variable tracking size limit exceeded with -fvar-tracking-assignments, retrying without” is NOT an error. It’s a NOTE, as the message says. The executable myServer will still be compiled successfully, just without var-tracking-assignments (which is only required for debugging anyway). There will be an executable myServer in your directory that you can run. I have already explained what this messages means and how you can deal with it, if it bothers you.
[1] https://open62541.org/doc/current/building.html#select-build-artefacts
[2] https://www.kalycito.com/running-opcua-on-raspberry-pi/
[3] https://reference.opcfoundation.org/v104/Core/docs/Part3/8.2.2/
Hello Benedict,
Thanks for your feedback and
sorry about the late reply.
Have now put the Raspberry back on again. Everything rebuilt and the gcc command executed with the -flto = 1 command. Now the whole thing can be built and executed with namespace Zero Full. However, it takes a good hour to build. A lecturer gave me the tip with crosscompiling that I would look at that next. Because the existing Raspberrys in the demonstrator are older and slower. Do you have experience with that I have to pay attention to what?
I am currently writing all of my results in a small manual, open62541 on raspberry for dummies 😉 I can make these available to you. In case you have to give tips to beginners like me 😉
Hi,
happy to hear that -flto=1 did the trick, congratulations on your success and thanks for reporting back your results.
Regarding cross-compiling:
1) Before setting up a cross-compiler try to do as much work as possible on your development computer or some spare computer/laptop
2) Setting up a cross-compiler requires some trial-and-error in my experience. I recommend you try following the instructions given by Stefan [1], who is also owner of this blog. Depending on your time budget setting up a cross-compiler toolchain might be not worth the effort.
3) Thanks to ‘make’, the build process will be faster if you’re rebuilding the code a second time with minor changes. How much time does a rebuild take?
Regarding your manual:
That’s very kind of you! I’ll contact you by mail to work something out.
Cheers!
[1] https://stackoverflow.com/questions/19162072/how-to-install-the-raspberry-pi-cross-compiler-on-my-linux-host-machine/58559140#58559140
Hello Benedict,
I’m already two steps further along with the crosscompiler, but it’s more likely to be a paid add-on for Visual Studio. I am currently testing VisualGDB here.
Then I had now also tried whether the whole compiles faster when I build it again. Unfortunately, nothing changes here, of course in relation to the make -j command, but not with the gcc command… that would have surprised me, the gcc command has no temporary files, which would make it faster. However, I am still a beginner and can also be completely wrong here.
Greetings Klemens
Hi Klemens,
good to see you making progress & thanks for sharing!
@ paid add-on for VS / VisalGDB:
Hm, I am not exactly sure what code you’re working on / debugging right now but to me it looks like VisualGDB reproduces many features in Visual Studio that you would find for free in any common Linux IDE. But whatever floats your boat, I don’t know your skills, preferences, setup, … 🙂
@ faster re-compiles:
You’re right. I had ‘make’ in mind. For a single gcc call it won’t help.
Cheers!
“… You can also use the following value instead: $HOME/install”
Nope. Does not work on Linux (at least on Debian), as the CMake Config-GUI will wrap $Home into a string. This results in having “make install” put everything into PATH/To/build/’$Home’/install
Anyone reading this comment, please read the follow-up comment at https://opcua.rocks/step-9-using-a-custom-uanodeset-with-open62541/ before jumping to conclusions.
Hi Benedict,
I was using your installation recommendation (https://opcua.rocks/step-1-setup-open62541-on-linux-build-first-server/) on ubuntu 16.04 LTS. And Basic Server implementation was working smoothly and fast. It was old computer which has 4GB totall ram size.
Now with new computer, i have Ubuntu 20.04.2 LTS and i reinstalled open62541 here.
And following code compilation takes at least 20 min with far better,faster PC
gcc -std=c99 -I$HOME/install/include -L$HOME/install/lib myServer.c -lopen62541 -lmbedtls -lmbedx509 -lmbedcrypto -o myServer
And i got the similar warnings as other colleagues mentioned before as below;
/home/osboxes/open62541/build/src_generated/open62541/namespace0_generated.c: In function ‘namespace0_generated’:
/home/osboxes/open62541/build/src_generated/open62541/namespace0_generated.c:118113:15: note: variable tracking size limit exceeded with ‘-fvar-tracking-assignments’, retrying without
118113 | UA_StatusCode namespace0_generated(UA_Server *server) {
| ^
However ./myServer is running without problem. What is the problem here about variable tracking size limit ?Why this compilation is slower with this ubuntu version?
I would be appreciated if u have any comment or idea…
Hi,
1) Did you verify that you’re using the same version / release / git commit on your old and new ubuntu? Use
git rev-parse HEADto compare git id.2) Did you verify that you’re using the same make configuration on your old and new ubuntu? Use
cd ~/open62541/build && ccmake ..to compare.3) Did you try your command with
-fvar-tracking-assignmentsas suggested by the compilation note message?4) Did you compare gcc versions? Use
gcc --versionWhat is the problem here about variable tracking size limit?
As explained in several(!) other comments: It is a feature of gcc to keep book of variables for better debugging. As long as you don’t plan to debug open62541 itself (I assume you don’t) you can disable it without any disadvantage to speed up compilation. Also, this message is just a note by the compiler for your information. It is not a warning or an error.
Why this compilation is slower with this ubuntu version?
Probably because you did not check the four points I have listed in this comment above.
Kind regards & good luck
Hi,
i’m just new to this whole topic. I just learned that this open62541 exists and started to mess around with the tutorial. An installation was given, although without any crypto modules. Compiling and running the server worked out of the box. So far so good. Running the server gave some warnings about missing crypto features, but that’s ok for now.
Now i’d like to continue with step 2. But here i do not have any docker software. Is it possible to compile the tutorial model on linux without docker?
I’m quite new to the whole concept. We have here drilling and routing machines, and software for it too. I did most of the communication features myself. Classical request-response. Built on xml and/or json. Now the request is: Do this same thing with open62541. And currently i do not see how this fits. So please forgive me if these are silly questions, i am still struggling to understand all this 🙂
Regards
Siegfried Gipp
Hi Siegfrid,
welcome to OPC UA!
The learning curve for OPC UA modeling and the tools surrounding it is steep. That’s why this tutorial was published. All questions welcome 🙂
Regarding your question: “Is it possible to compile the tutorial model on linux without docker?”
Answer: Have a look at the graph captured “Overview of available workflows and repository dependencies” in the intro article of this series at:
https://opcua.rocks/from-modelling-to-execution-opc-ua-information-model-tutorial/
This graph (and the text following it) show, that using docker on linux is optional and is provided for your convenience. If you use docker (the bold path on the very left of the graph), the UAModelCompiler comes ready to use. If you do not use docker, you will have to build the UAModelCompiler yourself, before you can compile any models. Depending on your software development experience on Linux, this may be more or less time consuming, since UAModelCompiler is a C# VisualStudio project with windows as targeted development platform. Making it run on linux requires mono. It’s possible (as the docker image proves) but it’s cumbersome too.
Important note: Please be aware that compiling the Model with UAModelCompiler will just transform den XML UAModel to XML UANodeSet2. After that, the UANodeSet2 will be used to compile the open62541 binary which runs on linux. Instead of having your complete toolchain on linux (as described above) you may also set up your tool chain as follows:
Develop and compile model on your Windows development machine with UAModelCompiler. Send the compiled model (UANodeSet2 XML) to your linux development machine and compile open62541.
Hope that helps.
Good luck & do not hesitate with follow-up questions!