aoc compilation fails: Error: save_system kernel_system.qsys (CENTOS 7)

[Altera_Forum]

Hi, 

 A colleague and I are having issues compiling the fft1d example provided by Intel/Altera. OCLSDK passes first stage compilation fine, but fails with the message 'Error: save_system kernel_system.qsys: 0'. ( Compiling on windows doesn't appear to be an issue but we would like to compile on our Centos server )  I've attached the .log, I think the last 3 lines are relevant -   2018.03.23.10:02:26 Info: save_system kernel_system.qsys 2018.03.23.10:02:26 Info: Replacing kernel_system.clk_1x with generic component 2018.03.23.10:02:26 Error: save_system kernel_system.qsys: 0  My belief is that somehow that replacing the kernel_system.clk_1x is causing save_system to fail. I've googled around and there isn't a lot online relating to the save_system issue.   Thanks for your time and any advice that can be offered.  ap29

[Altera_Forum]

Please post output of "aoc --version" and "quartus_map --version" and attach the "quartus_sh_compile.log" file.

[Altera_Forum]

--- Quote Start ---  Please post output of "aoc --version" and "quartus_map --version" and attach the "quartus_sh_compile.log" file.  --- Quote End ---    Hi HRZ,  okay, will get these logs ASAP

[Altera_Forum]

--- Quote Start ---  Please post output of "aoc --version" and "quartus_map --version" and attach the "quartus_sh_compile.log" file.  --- Quote End ---    Output of aoc --version is:  Intel(R) FPGA SDK for OpenCL(TM), 64-Bit Offline Compiler Version 16.1.2 Build 203 Copyright (C) 2017 Intel Corporation  Output of quartus_map --version:  Quartus Prime Analysis & Synthesis Version 16.1.2 Build 203 01/18/2017 SJ Pro Edition Copyright (C) 2017 Intel Corporation. All rights reserved.  There does not appear to be a quartus_sh_compile.log file for fft1d - we have found this log for other projects only such as library example 1 and vector_add.  I've also copied some of the terminal output:  > aoc --board de5a_net_e1 device/fft1d.cl -o bin/fft1d.aocx -v aoc: Environment checks are completed successfully. You are now compiling the full flow!! aoc: Selected target board de5a_net_e1 aoc: Running OpenCL parser.... aoc: OpenCL parser completed successfully. aoc: Compiling.... aoc: Linking with IP library ... aoc: First stage compilation completed successfully. Error: Qsys-script FAILED. Refer to fft1d/fft1d.log for details.  Thanks for your time, and any help you can provide,  ap29

[Altera_Forum]

So, you have successfully compiled other code on the same machine? If "quartus_sh_compile.log" is not generated, then it means compilation is failing very early in the process. Do you have enough free space on your disk? Also, have you tried removing the existing compilation folder for fft1d and retrying the compilation?

[Altera_Forum]

--- Quote Start ---  So, you have successfully compiled other code on the same machine? If "quartus_sh_compile.log" is not generated, then it means compilation is failing very early in the process. Do you have enough free space on your disk? Also, have you tried removing the existing compilation folder for fft1d and retrying the compilation?  --- Quote End ---    Disk space isn't an issue (200 GB spare). Have tried removing the compilation folder and recompiling, but the results is the same QSYS-script FAILED error message. Tried compiling another example (vector_add) and it failed in the same way. I guess the quartus_sh_compile.log must be generated from the files when we were compiling on windows, and when we copied the examples across, we must have copied the logs too. Any other ideas?

[Altera_Forum]

Without the "quartus_sh_compile.log" file it is hard to speculate what is wrong. You can try reinstalling Quartus and your OpenCL BSP and try again. Some people previously reported problems with Terasic's 16.1 BSP for the DE5a-Net board on Linux and I confirmed the issue on my own environment: 

 https://www.alteraforum.com/forum/showthread.php?t=56560&p=230072#post230072 http://www.alteraforum.com/forum/showthread.php?t=56531  However, your problem seems to be different from that case and I came across at least one person in the forum who claimed Terasic's BSP works fine for him.