Author: documenteddesigns

Building Kicad with QT Creator on Windows/Msys2

I have been working towards integrating RenumKicadPCB functionality (i.e. geographic back-annotation) into Kicad for some time. Actually, to be accurate I have been hoping to do that. A major roadblock has been an IDE which allows me to edit, build, and debug the code. I had experimented with Eclipse and Microsoft Visual Studio Community Edition but I couldn’t import the official cmake files as there was always something missing. I am, after all, a former hardware guy not a modern programmer.

Recently back-annotation came up in a Kicad developer’s thread and I made contact with Alexander Shuklin, who is not only a more advanced software guy but he has already contributed to Kicad. We agreed to collaborate on the project – a development which I am very happy with. Alexander uses QT-Creator and, while I was initially hoping he could help me get Kicad working on MSVC I realized it would probably be easier for me to get QT-Creator working since at least I know you can build Kicad on that IDE.

After several days of failure interspersed with a near success (I was so tired I forgot to document what I had done) I managed to get things going. Here is the procedure to build Kicad on Windows with QT-Creator. Note that this only seems to work with the Msys2 version of QT-Creator, not the Windows version of QT-Creator even if Msys2 is installed. Unfortunately this took a day and a half to figure out. I almost succeed with the QT-Creator on Windows (I actually got about 33% through the compile) but, as I mentioned I forgot to document what I done.

First install Msys2 (see http://repo.msys2.org/distrib/x86_64/msys2-x86_64-20161025.exe )

Everything below is from a mingw64 terminal window. It does not seem to work if you pick from the Msys2 pop-up. I show 64 bit development because that is what is recommended. Similar steps would apply to a 32 bit development environment.

There may be redundant or unnecessary steps in this recipe. I completely deleted Msys2 a few times and re-ran this procedure but there are only so many hours in a day.

From the Mingw64 terminal (comments marked by //)

pacman -Syuu //Answer yes to remove conflicts close the window via task manager when done

pacman -Syuu //Takes a long time close the window via task manager when done

pacman -Syuu //To make sure all is up to date

Copy and paste the following lines into your Mingw64 terminal (all the way from pacman -S to the blank line)

pacman -S base-devel \
git \
mingw-w64-x86_64-cmake \
mingw-w64-x86_64-doxygen \
mingw-w64-x86_64-gcc \
mingw-w64-x86_64-python2 \
mingw-w64-x86_64-pkg-config \
mingw-w64-x86_64-swig \
mingw-w64-x86_64-boost \
mingw-w64-x86_64-cairo \
mingw-w64-x86_64-glew \
mingw-w64-x86_64-curl \
mingw-w64-x86_64-wxPython \
mingw-w64-x86_64-wxWidgets \
mingw-w64-x86_64-toolchain \
mingw-w64-x86_64-glm \
mingw-w64-x86_64-oce \
mingw-w64-x86_64-ngspice
mercurial \
cvs \
p7zip \
ruby \
mingw-w64-i686-toolchain mingw-w64-x86_64-toolchain

Now copy these lines one at a time (it may be possible to combine these with the above)

pacman -S mingw-w64-i686-qt-creator mingw-w64-x86_64-qt-creator

pacman -S mingw-w64-x86_64-clang

pacman -Syuu

Run qtcreator to make sure it works then exit qtcreator.

qtcreator

Download source files eg kicad-5.1.4.tar.xz into a suitable location such as /home source files can be found at http://kicad-pcb.org/download/source/ then

tar -xaf kicad-5.1.4.tar.xz

Run qtcreator from the ming64 terminal (for some reason, it doesn’t work if you pin it to the taskbar)

qtcreator

Select Open File or Project and open the Kicad source directory. Wait for it to parse

Note that QT will make its own build directory. This doesn’t seem to matter.

Select Build, rebuild project. On my machine it takes about 2 hours to compile Kicad

I have fixed a bug found by Paul P. where V0410 would not run properly if there was no existing RenumKiCadPCB directory in /users/appdata/roaming (Thanks Paul!).

Also at his suggestion I now show a warning and do not save the updated netlist file if there are netlist errors. This is usually because the netlist is out of date with the schematic and PCB so it makes it easier to fix (just regenerate the netlist from the schematic).

Here’s the link on Github https://github.com/BrianAtDocumentedDesigns/RenumKiCadPCB0401_July12_2019/tree/master

 

Here it is … https://github.com/BrianAtDocumentedDesigns/RenumKiCadPCBo410

Version 410 has been significantly rewritten as I move towards KiCad integration.
On the journey some bugs were reported (Thanks Alan!)

These bugs which were reported and fixed are:
1) Under some circumstances KiCad uses quotes around reference designations internally. This caused RenumKiCadPCB to treat each refdes as a different type.

2) Renumbering by refdes didn’t work.

3) Selecting “Sort by Refdes” wasn’t remembered after exiting the program even though “Sort by Modules”was.

Also I added a reminder warning that you have to close other versions of the relevant .sch and .kicad_pcb files before renumbering. KiCad doesn’t refresh externally modified files and I didn’t see the benefit of sorting out file lock on Windows, Linux, and Apple given I hope to integrate the code into KiCad and therefore the issue will be moot.

Edit: I rather stupidly forgot the link to the new version ….

https://github.com/BrianAtDocumentedDesigns/RenumKiCadPCBo410

 

I have entirely re-written RenumKiCadPCB into c++ and made it into a GUI (Windows-like) program. All the source files as well as build instructions and a Windows 10 installation file (RenumKiCadPCB.msi) can be found at https://github.com/BrianAtDocumentedDesigns/RenumKiCadPCB0400

I also added a few features and fixed a number of bugs.

As usual, please contact brian@documenteddesigns.com with bug reports or feature requests.

 

 

I just uploaded V0.352 to Github. This (hopefully) fixed a bug where the system could chose the wrong “Top Layer”. It also (hopefully) corrected an issue with the undocumented AR field in the schematic. The field is not only undocumented, as implemented it can lead to “ghost” components in the file which are not visible to eeSchema or PCBNew. This, in turn led to errors in renumbering which could previously only be corrected by manually editing the schematic file with a text editor.

Although I have tested the code on my system I have not received feedback with respect to whether it works (or builds) for others.

I am re-writing RenumKiCadPCB from scratch in c++. I am doing so in order to learn c++ and wxWidgets and then create a GUI version of the program. Doing so may provide enough of a grounder to produce a version of eeSchema and/or PCBNew with integral renumbering. It should be quite easy to do this once I get the hang of a few parts of the KiCad source. The developers would like to see renumbering “pushed” from PCB to schematic via a netlist update. This probably makes sense in the broader context of things like pin swapping but since no progress has been mode in this direction it will have to work.

I intend to add a number of features to V0.400 (which I might release before V0.500, the GUI version). These are:

1. Enable global change to module text angle;
2. Enable global change to module text location (i.e. align Value and/or User with Ref, center the text fields on the module;
3. Generate a Was/Is file from the PCB;
4. Update schematic only using a Was/Is file.

The last two items may seem odd but it looks like updating the schematic only and then using eeSchema to update the PCB using timestamps regenerates the PCB and the netlist directly using KiCad internal functions so this would be a stepping stone to full integration, or at least less likelihood of RenumKiCadPCB getting out of sync with the file formats which appear to evolve and not be well documented. The net result of was/is generation followed by update PCB is substantially identical to pushing a netlist from PCB to schematic.

I would greatly appreciate any feature suggestions, bug reports, or general encouragement.

 

When I started having trouble with WordPress and zip files I moved the repository for RenumKicadPCB to DropBox. I recently was alerted that for some I had to give a user “permission” to access the link even though it was supposed to be shared. I have no problem with this but I understand some users might so I decided to move the project to Github. Here is the link https://github.com/BrianAtDocumentedDesigns/RenumKicadPCB

 

I completed a rewrite a couple months but didn’t do all the testing I wish I had.

I’ve found no problems with this code but admit I have been focusing on Kicad proper with the hope the developers will add renumbering to V6.

Let me know if there are issues.

For some reason WordPress is blocking me from using zip files. Here is a link to the file on Dropbox

https://www.dropbox.com/s/jekf97odywc5bdx/RenumKiCadPCBV0351.zip?dl=0

Direct link (no log in) View file

 

 

Thanks to Michael I learned that there are issues with at least the two most recent exe files zip files I posted.

Somehow I was distributing without a dll which can be corrected by compiling like this under msys2

gcc RenumKiCadPCB.c -static-libgcc -static-libstdc++ then renamining a.exe to renum.exe (or whatever).

Here is the corrected file RenumKiCadPCBV0300-fixed

Sorry about the screw up. I had been trying to compile Kicad sources and must have modifile my compiler settings or something.

There appears to be a bug I thought I had fixed, namely that a windows file path with a “.” (i.e. ver1.c.pro) gets trimmed wrongly. I will update soon.

 

V0.300 RenumKiCadPCBV0300-fixed  (NOTE: THIS IS AN UPDATED ZIP WITH A FIXED EXE FILE) is a substantial rewrite in order to process errors better and to fix bugs which popped up in stress testing.

User Jano reported “The make RenumKiCadPCB don’t works out of the box, instead I was able to compile with: gcc RenumKiCadPCB.c -o RenumKiCadPCB -lm”

I don’t know if this is an issue across Linux versions but, just in case I have updated the manual in the zip.

Now everything is done in memory before writing to files.  If there are warnings, it asks whether to commit changes. After all, some errors, like mounting holes on the PCB which are not on the schematic, are not critical, even though you really should fix errors (including ERC and DRC).

If there are no errors (or if the user decides to “commit”), all related files are backed up and the new ones are written.

I fixed a number of issues associated with/exposed by weird file content in TERES_PCB1-A64-MAIN_Rev.C.pro.

I have stress tested with https://github.com/OLIMEX/DIY-LAPTOP/tree/master/HARDWARE/A64-TERES/TERES-PCB1-A64-MAIN_Rev_C which is the most complex KiCad project I have come across.

The Olimex Rev C files themselves have issues:

ERC and DRC show errors. I was able to fix all the ERC errors:
1) Resistor networks in the schematic have sections with different modules than                     the PCB
2) There is an excess PWR_FLAG on a power connector.

Also (bug reported to the developers) LCD.SCH has a “phantom” instantiation which does  not show up in ERC. The phantom instantiation meant parts were labeled U? but didn’t show up as not annotated. Even though renumKiCadPCB works, it flags errors such as this.

There are blank reference designations (i.e. “”) on the PCB.

The good news is, all these errors exposed issues with RenumKiCadPCB!

The Olimex PCB fails DRC with 9 errors as distributed. After creation and importation of a netlist this becomes 27 errors. The new errors appear to be associated with zones, a well established KiCad bug (see MANUAL.TXT).

I couldn’t figure out Olimex’s license, otherwise I’d post the fixed files (renumbered!)

After RenumKiCadPCB is run on the original files or the fixed files the DRC errors are the same, respectively. I assume, therefore, that RenumKiCadPCB is working. (famous last words …)

The Kicad “pour error” described in MANUAL.TXT still seems to be present in Kicad V5 nightly builds. The way you fix it (in V5 nightlies) is to “Edit” the zone (which I find a pain in the butt to do – type “E” exactly on the edge zone outline).

Make sure Net Filtering Display: Show All is selected and rename the zone net.

I remain hopeful the developers will decide to incorporate RenuKiCadPCB functionality into the software. It should be easy to do (but I don’t know how). Again, if there is anybody out there willing to do encourage the KiCad development group, please do.

Similarly, I would welcome anybody with requisit c++ or GUI programming skills to put a wrapper on the code.

PLEASE LET ME KNOW OF ANY BUGS!