I'm using the build.py script from the Mathworks buildroot repository at https://github.com/mathworks/buildroot
I'm running the following commands:
./build.py -b zc706 -p zynq
./build.py -b zcu102 -p zynqmp
Since we want to use MATLAB with the zcu102 station, we'll focus on that one.
The first problem was a missing locale.h.
It looks like the custom glibc (so that Simulink will run on Ubuntu 18.04) is preventing buildroot from finding locale.h
Fortunately we had another Ubuntu 18.04 machine without the custom glibc, so we moved over there.
We use Ubuntu 18.04 because we're using Vivado 2020.2. Ubuntu 20.04 and later do not officially support Vivado.
Vivado 2020.2 has been successfully tested with MATLAB 2022b and HDL Coder. Using HDL Coder is the goal.
After moving away from the custom glibc required to run Simulink, we got another error. It is below:
Cannot execute cross-compiler '/opt/linaro/aarch32-6.3.1-2017.02/bin/arm-linux-gnueabihf-gcc'
package/pkg-generic.mk:217: recipe for target '/home/abraxas3d/haifuraiya/buildroot/output/zc706_linux_linaro/build/toolchain-external-custom/.stamp_configured' failed
make[1]: *** [/home/abraxas3d/haifuraiya/buildroot/output/zc706_linux_linaro/build/toolchain-external-custom/.stamp_configured] Error 1
Makefile:16: recipe for target '_all' failed
make: *** [_all] Error 2
Uncaught exception
Traceback (most recent call last):
File "./build.py", line 191, in <module>
build_target(args, catalog)
File "./build.py", line 89, in build_target
subproc(argStr, cwd=args['outputDir'])
File "/home/abraxas3d/haifuraiya/buildroot/board/mathworks/common/scripts/helper_func.py", line 492, in subproc
_SUBPROC_LOGGER.call(args, cwd=cwd, shell=shell)
File "/home/abraxas3d/haifuraiya/buildroot/board/mathworks/common/scripts/helper_func.py", line 77, in call
raise subprocess.CalledProcessError(self.proc.returncode, self.strArgs(args), None)
CalledProcessError: Command 'make legal-info all' returned non-zero exit status 2
So, we couldn't execute the cross-compiler. The cross-compiler wasn't in this location, which appears to be hard-coded into build.py when you do not call out a custom location for the cross-compiler.
Installing the cross compilier from the package manager in Ubuntu, and pointing to it's location in usr/bin/ gave a different specific error, that package installations of a cross-compiler could not be used by buildroot.
Copying over files from the package install to opt/ did not work.
Finding the aarch32-6.3.1-2017.02 tarball from Linaro, and expanding it into opt/ and renaming the directories to match what build.py wanted got things much further along.
On Ubuntu 18.04, using the tag R22.2.0, and trying to build for the zc706, this build failed due to a problem with e2fsprogs.
./../misc/create_inode.c:395:18: error: conflicting types for 'copy_file_range'
static errcode_t copy_file_range(ext2_filsys fs, int fd, ext2_file_t e2_file,
^~~~~~~~~~~~~~~
In file included from ./../misc/create_inode.c:19:0:
/usr/include/unistd.h:1110:9: note: previous declaration of \u2018copy_file_range\u2019 was here
ssize_t copy_file_range (int __infd, __off64_t *__pinoff,
^~~~~~~~~~~~~~~
SUBST debugfs.8
Makefile:423: recipe for target 'create_inode.o' failed
make[4]: *** [create_inode.o] Error 1
make[4]: *** Waiting for unfinished jobs....
Makefile:405: recipe for target 'all-progs-recursive' failed
make[3]: *** [all-progs-recursive] Error 1
Makefile:340: recipe for target 'all' failed
make[2]: *** [all] Error 2
package/pkg-generic.mk:227: recipe for target '/home/abraxas3d/haifuraiya/buildroot/output/zc706_linux_linaro/build/host-e2fsprogs-1.43.4/.stamp_built' failed
make[1]: *** [/home/abraxas3d/haifuraiya/buildroot/output/zc706_linux_linaro/build/host-e2fsprogs-1.43.4/.stamp_built] Error 2
Makefile:16: recipe for target '_all' failed
make: *** [_all] Error 2
Uncaught exception
Traceback (most recent call last):
File "./build.py", line 191, in <module>
build_target(args, catalog)
File "./build.py", line 89, in build_target
subproc(argStr, cwd=args['outputDir'])
File "/home/abraxas3d/haifuraiya/buildroot/board/mathworks/common/scripts/helper_func.py", line 492, in subproc
_SUBPROC_LOGGER.call(args, cwd=cwd, shell=shell)
File "/home/abraxas3d/haifuraiya/buildroot/board/mathworks/common/scripts/helper_func.py", line 77, in call
raise subprocess.CalledProcessError(self.proc.returncode, self.strArgs(args), None)
CalledProcessError: Command 'make legal-info all' returned non-zero exit status 2
The tagged branch in the repository calls out e2fsprogs version 1.43.4. However, this version of e2fsprogs has a function name collision. That is what the error: conflicting types for 'copy_file_range' is indicating. This was fixed in e2fsprogs version 1.43.8.
The next step was to modify e2fsprogs.mk file from mathworks/buildroot repository and attempt to build the SD Card image again.
And this is how to patch it:
Create a .patch file and put it in the buildroot/package/e2fsprogs folder
At this point we changed to MATLAB 2021b because this is the version that Analog Devices Transceiver Toolbox was compatible with. This meant checking all the other toolchains for compatibility.
"MATLAB R2021b HDL Coder has been tested against Xilinx Vivado 2020.1"
We sourced this version of Vivado and proceeded.
However, from the Transceiver Toolbox github release page, we noticed:
v21.2.1 Version
Upgraded to support R2021b and significant cleanup of HDL infrastructure.
Supported Tool Versions:
- MATLAB R2021b
- Xilinx Vivado 2019.1
- 2019-R2 SD Card (Kuiper Linux)
So, which Vivado do we need to use? This makes a big difference. For example, which branch of the HDL Reference design do we use? Is 2019.1 compatible wtih HDL Coder?
"MATLAB R2020a has been tested against Xilinx Vivado 2019.1"
So, do we back off to R2020a? Or, will this break something else? We decided to stay with R2021b, but Transceiver Toolbox not being compatible with the Vivado that R2021b HDL Coder has been tested against seems like an obvious risk.
New error:
-- Check for working C compiler: /home/abraxas3d/haifuraiya/buildroot/output/zc706_linux_linaro/host/usr/bin/arm-linux-gnueabihf-gcc -- works
-- Detecting C compiler ABI info
-- Detecting C compiler ABI info - done
-- Detecting C compile features
-- Detecting C compile features - done
-- Found PkgConfig: /home/abraxas3d/haifuraiya/buildroot/output/zc706_linux_linaro/host/usr/bin/pkg-config (found version "0.28")
-- Detected version of GNU GCC: 63 (603)
-- Looking for ccache - found (/usr/bin/ccache)
CMake Error at cmake/OpenCVCompilerOptions.cmake:21 (else):
A duplicate ELSE command was found inside an IF block.
Call Stack (most recent call first):
CMakeLists.txt:432 (include)
-- Configuring incomplete, errors occurred!
See also "/home/abraxas3d/haifuraiya/buildroot/output/zc706_linux_linaro/build/opencv-2.4.13.2/CMakeFiles/CMakeOutput.log".
package/pkg-generic.mk:217: recipe for target '/home/abraxas3d/haifuraiya/buildroot/output/zc706_linux_linaro/build/opencv-2.4.13.2/.stamp_configured' failed
make[1]: *** [/home/abraxas3d/haifuraiya/buildroot/output/zc706_linux_linaro/build/opencv-2.4.13.2/.stamp_configured] Error 1
Makefile:16: recipe for target '_all' failed
make: *** [_all] Error 2
Uncaught exception
Traceback (most recent call last):
File "./build.py", line 191, in <module>
build_target(args, catalog)
File "./build.py", line 89, in build_target
subproc(argStr, cwd=args['outputDir'])
File "/home/abraxas3d/haifuraiya/buildroot/board/mathworks/common/scripts/helper_func.py", line 492, in subproc
_SUBPROC_LOGGER.call(args, cwd=cwd, shell=shell)
File "/home/abraxas3d/haifuraiya/buildroot/board/mathworks/common/scripts/helper_func.py", line 77, in call
raise subprocess.CalledProcessError(self.proc.returncode, self.strArgs(args), None)
CalledProcessError: Command 'make legal-info all' returned non-zero exit status 2
Zeroing in on the error message:
-- Found PkgConfig: /home/abraxas3d/haifuraiya/buildroot/output/zc706_linux_linaro/host/usr/bin/pkg-config (found version "0.28")
-- Detected version of GNU GCC: 63 (603)
-- Looking for ccache - found (/usr/bin/ccache)
CMake Error at cmake/OpenCVCompilerOptions.cmake:21 (else):
A duplicate ELSE command was found inside an IF block.
Call Stack (most recent call first):
CMakeLists.txt:432 (include)
This seems to be related to a bug reported here:
System information (version)• OpenCV => 2.4.13.3
• Operating System / Platform => ArchLinux 64-bit
• Compiler => gcc 7.1.1 with ccache 3.3.4 installed
Detailed descriptionOnce I run cmake, it produced the following error.
-- Found PkgConfig: /usr/local/bin/pkg-config (found version "0.29.2")
-- Looking for ccache - found (/usr/local/bin/ccache)
CMake Error at cmake/OpenCVCompilerOptions.cmake:21 (else):
"Fixed in 2.4 and master branch"
As far as I can tell, this fix was supposed to be backported to version 2.4.
According to the opencv.mk file, we have OPENCV_VERSION = 2.4.13.2
OpenCV => 2.4.13.3 was the version mentioned in the github issue about the misplaced "else" command.
So, our OpenCV called in by Mathworks Buildroot could be too old to get that bug fix.
I edited the opencv.mk file to update the version of opencv, like I did for e2fsprogs. The release page for OpenCV says this version is from 2014, so I'm somewhat concerned about how old this is. Not that there's anything wrong with old code. Mathworks buildroot is using a Linaro cross compiler for ARM from 2017. Overall, though, these sorts of errors in the current Matlab linux build for Zynq dev boards makes me really nervous. We're using Ubuntu 18.04 here because that's what Vivado will run on.I'm wondering if this repository ever been tested or used on Ubuntu 18.04?
I'm edited the OpenCV.hash file and rerun the build.
Checking what versions of OpenCV had this particular commit:
abraxas3d@chococat:~/Downloads/opencv$ git tag --contains 126de0cd95418811e302996161e03195bf7e631a
2.4.13.4
2.4.13.5
2.4.13.6
2.4.13.7
Going with 2.4.13.4
The commit ID above is the one that fixes the cmake error with misplaced else statement.
Downloaded 2.4.13.4 tarball from github (opencv)
abraxas3d@chococat:~/Downloads$ sha256sum opencv-2.4.13.4.tar.gz
f8abf1fcc2da3bb1deac8776f07b8390f871372e2a44dc355c765dd379194481 opencv-2.4.13.4.tar.gz
updated hash in opencv.hash with the locally generated hash.
next error is also in opencv
Hunk #1 succeeded at 409 (offset 5 lines).
patching file modules/ocl/src/matrix_operations.cpp
patching file modules/python/src2/cv2.cpp
Applying 0002-avoid-sysctl_h.patch using patch:
patching file modules/core/src/parallel.cpp
Reversed (or previously applied) patch detected! Skipping patch.
1 out of 1 hunk ignored -- saving rejects to file modules/core/src/parallel.cpp.rej
patching file modules/core/src/system.cpp
Hunk #1 FAILED at 163.
1 out of 1 hunk FAILED -- saving rejects to file modules/core/src/system.cpp.rej
patching file modules/highgui/src/cap_ffmpeg_impl.hpp
Reversed (or previously applied) patch detected! Skipping patch.
1 out of 1 hunk ignored -- saving rejects to file modules/highgui/src/cap_ffmpeg_impl.hpp.rej
package/pkg-generic.mk:191: recipe for target '/home/abraxas3d/haifuraiya/buildroot/output/zc706_linux_linaro/build/opencv-2.4.13.4/.stamp_patched' failed
make[1]: *** [/home/abraxas3d/haifuraiya/buildroot/output/zc706_linux_linaro/build/opencv-2.4.13.4/.stamp_patched] Error 1
Makefile:16: recipe for target '_all' failed
make: *** [_all] Error 2
Uncaught exception
Traceback (most recent call last):
File "./build.py", line 191, in <module>
build_target(args, catalog)
File "./build.py", line 89, in build_target
subproc(argStr, cwd=args['outputDir'])
File "/home/abraxas3d/haifuraiya/buildroot/board/mathworks/common/scripts/helper_func.py", line 492, in subproc
_SUBPROC_LOGGER.call(args, cwd=cwd, shell=shell)
File "/home/abraxas3d/haifuraiya/buildroot/board/mathworks/common/scripts/helper_func.py", line 77, in call
raise subprocess.CalledProcessError(self.proc.returncode, self.strArgs(args), None)
CalledProcessError: Command 'make legal-info all' returned non-zero exit status 2
We look at these rejection files to see what's going on.
abraxas3d@chococat:~/haifuraiya/buildroot/output/zc706_linux_linaro/build/opencv-2.4.13.4/modules/core/src$ cat parallel.cpp.rej
--- modules/core/src/parallel.cpp
+++ modules/core/src/parallel.cpp
@@ -56,7 +56,7 @@
#include <sys/types.h>
#if defined ANDROID
#include <sys/sysconf.h>
- #else
+ #elif defined __APPLE__
#include <sys/sysctl.h>
#endif
#endif
abraxas3d@chococat:~/haifuraiya/buildroot/output/zc706_linux_linaro/build/opencv-2.4.13.4/modules/core/src$ cat system.cpp.rej
--- modules/core/src/system.cpp
+++ modules/core/src/system.cpp
@@ -163,7 +163,7 @@ std::wstring GetTempFileNameWinRT(std::wstring prefix)
#include <sys/types.h>
#if defined ANDROID
#include <sys/sysconf.h>
-#else
+#elif defined __APPLE__
#include <sys/sysctl.h>
#endif
#endif
No idea why Android and Apple support being added would cause this buildroot fail. We need some help figuring this out.
same error with 2.4.13.5
same error with 2.4.13.6
same error with 2.4.13.7, which is the end of version 2.4 of OpenCV.
-=-=-=-=-=-=-=-=-=-
We switched to a third machine and installed Vivado 2020.2
This OS is an Ubuntu 20.04 (Vivado 2020.2 is listed as supported by Ubuntu 20.04) with the R21.2.0 branch of mathworks buildroot.
After manually installing the ARM Linaro cross compiler in /opt/linaro and renaming the directory to match the default in the build.py script, the following error happened:
freadahead.c: In function 'freadahead': freadahead.c:92:3: error: #error "Please port gnulib freadahead.c to your platform! Look at the definition of fflush, fread, ungetc on your system, then report this to bug-gnulib." 92 | #error "Please port gnulib freadahead.c to your platform! Look at the definition of fflush, fread, ungetc on your system, then report this to bug-gnulib." | ^~~~~ make[5]: *** [Makefile:1915: freadahead.o] Error 1 make[5]: *** Waiting for unfinished jobs.... CC fseeko.o fseeko.c: In function 'rpl_fseeko': fseeko.c:110:4: error: #error "Please port gnulib fseeko.c to your platform! Look at the code in fseeko.c, then report this to bug-gnulib." 110 | #error "Please port gnulib fseeko.c to your platform! Look at the code in fseeko.c, then report this to bug-gnulib." | ^~~~~ make[5]: *** [Makefile:1915: fseeko.o] Error 1 make[4]: *** [Makefile:1674: all] Error 2 make[3]: *** [Makefile:1572: all-recursive] Error 1 make[2]: *** [Makefile:1528: all] Error 2
We backed up and found that there was a patch for the "misplaced else" OpenCV errors. It's commit c2e28if in OpenCV version 2.4 and can be created by "git format-patch -1 c2e281f"
With this patch, things moved forward.
Eventually an SD card image was created by using build.py. However it did not have a device tree for the ADRV9002.
We now needed to get the device tree for the ADRV9002 into Mathworks Buildroot linux and figure out how to compile it in.
In order to do this, the device tree source file from Analog Devices was obtained.
Where does this file go?
It turns out it goes in .../buildroot/board/mathworks/zynqmp/boards/zcu102/dts/linux/arch/arm64/boot/dts/xilinx/zynqmp-zcu102-rev10-adrv9002.dts
A set of errors happened. There are files included in the .dts that were missing. Here's an example:
ERROR:root: In file included from arch/arm64/boot/dts/zynqmp-zcu102-revA.dts:12:0,
ERROR:root: from arch/arm64/boot/dts/zynqmp-zcu102-revB.dts:10,
ERROR:root: from arch/arm64/boot/dts/zynqmp-zcu102-rev1.0.dts:10,
ERROR:root: from arch/arm64/boot/dts/zynqmp-zcu102-rev10-adrv9002.dts:10:
ERROR:root: arch/arm64/boot/dts/zynqmp.dtsi:15:47: fatal error: dt-/dma/xlnx-zynqmp-dpdma.h: No such file or directory
ERROR:root: #include <dt-bindings/dma/xlnx-zynqmp-dpdma.h>
The missing files were tracked down from the Analog Devices github by following the error messages.
abraxas3d@chococat:~/neptune/buildroot/board/mathworks/zynqmp/boards/zcu102/dts$ ls
adi-adrv9002.dtsi axistream.dts dt-bindings zynqmp-clk-ccf.dtsi zynqmp-zcu102-rev1.0.dts
axilite.dts base.dtsi linux zynqmp.dtsi zynqmp-zcu102-revA.dts
axistream_64.dts dlhdl.dts uapi zynqmp-zcu102-rev10-adrv9002.dts zynqmp-zcu102-revB.dts
When all the files were gathered, build.py errored out in the following way:
ERROR:root: /home/abraxas3d/neptune/buildroot/configs/zynqmp_zcu102_dynconfig_defconfig:189:warning: override: reassigning to symbol BR2_LINUX_KERNEL_DTS_SUPPORT
{ echo "ERROR: we shouldn't have a /etc/ld.so.conf file"; exit 1; } || true
{ echo "ERROR: we shouldn't have a /etc/ld.so.conf.d directory"; exit 1; } || true
ERROR:root: board/mathworks/zynqmp/scripts/postbuild_common.sh: illegal option -- i
ERROR:root: devicetree_axilite.dtb: ERROR (phandle_references): Reference to non-existent node or label "clkc"
Here's where we were stopped. We're not sure how to fix this yet or what the error message is trying to tell us.
-Abraxas3d
