Community Packages

Please read this before reporting a bug:
https://wiki.archlinux.org/title/Bug_reporting_guidelines

Do NOT report bugs when a package is just outdated, or it is in the AUR. Use the 'flag out of date' link on the package page, or the Mailing List.

REPEAT: Do NOT report bugs for outdated packages!
Tasklist

FS#62675 - [sdrangel] "illegal hardware instruction" crash

Attached to Project: Community Packages
Opened by sven (commonuser) - Saturday, 18 May 2019, 17:48 GMT
Last edited by Filipe Laíns (FFY00) - Friday, 06 December 2019, 09:31 GMT
Task Type Bug Report
Category Packages
Status Closed
Assigned To Filipe Laíns (FFY00)
Architecture All
Severity Low
Priority Normal
Reported Version
Due in Version Undecided
Due Date Undecided
Percent Complete 100%
Votes 0
Private No

Details

Description:
Crashes on startup, with and without SDR hardware and prints "illegal hardware instruction".

Additional info:

* Already tried to reinstall the package.

* Stacktrace:
Stack trace of thread 10301:
#0 0x00007f8154047fe9 _ZNSt8__detail8_ScannerIcEC2EPKcS3_NSt15regex_constants18syntax_option_typeESt6locale (libsdrbase.so)
#1 0x00007f815405571a _ZNSt8__detail9_CompilerINSt7__cxx1112regex_traitsIcEEEC2EPKcS6_RKSt6localeNSt15regex_constants18syntax_option_typeE (libsdrbase.so)
#2 0x00007f8154055ea8 _ZNSt8__detail13__compile_nfaINSt7__cxx1112regex_traitsIcEEPKcEENSt9enable_ifIXsrNS_27__is_contiguous_normal_iterIT0_EE5valueESt10shared_ptrIKNS_4_NFAIT_EEEE4typeES8_S8_RKNSC_11locale_typeENSt15regex_constants18syntax_option_typeE (libsdrbase.so)
#3 0x00007f8154056067 _ZNSt7__cxx1111basic_regexIcNS_12regex_traitsIcEEEC2EPKcNSt15regex_constants18syntax_option_typeE (libsdrbase.so)
#4 0x00007f8153fb2749 n/a (libsdrbase.so)
#5 0x00007f81542526aa call_init.part.0 (ld-linux-x86-64.so.2)
#6 0x00007f81542527aa _dl_init (ld-linux-x86-64.so.2)
#7 0x00007f815424403a _dl_start_user (ld-linux-x86-64.so.2)

* Demangled: std::__detail::_Scanner<char>::_Scanner(char const*, char const*, std::regex_constants::syntax_option_type, std::locale)

Steps to reproduce:
Start "sdrangel" from the console.
This task depends upon

Closed by  Filipe Laíns (FFY00)
Friday, 06 December 2019, 09:31 GMT
Reason for closing:  Fixed
Additional comments about closing:  sdrangel/sdrangel-sse4.1 4.12.1-1
Comment by Antonio Rojas (arojas) - Saturday, 18 May 2019, 17:59 GMT
should probably be built with -DNO_DSP_SIMD=ON
Comment by sven (commonuser) - Saturday, 18 May 2019, 18:20 GMT
Maybe also with -DENABLE_DISTRIBUTION=ON?

EDIT: Seems they only use -DENABLE_DISTRIBUTION=ON in their Debian build file.
Comment by Filipe Laíns (FFY00) - Saturday, 18 May 2019, 19:54 GMT
Seems using -DNO_DSP_SIMD=ON is the correct option, -DENABLE_DISTRIBUTION=ON still uses SIMD instructions.

https://github.com/f4exb/sdrangel/blob/9ab46c6056be3052dc63def1698a243389f2fff5/CMakeLists.txt#L160
Comment by sven (commonuser) - Saturday, 18 May 2019, 20:45 GMT
Is there a minimum CPU "version" for Arch? Maybe it is fine to use SSE3, which seems to be enabled by "ENABLE_DISTRIBUTION".

If I read the CMakeLists correctly, it will configure SSE by testing the host machine, except ENABLE_DISTRIBUTION is set. "NO_DSP_SIMD" seems only a preprocessor definition and may not prevent the compiler from using SSE in generated code.

Maybe we can patch the CMakeLists so "ENABLE_DISTRIBUTION" sets the SSE version according the minimum supported version by Arch?

EDIT: Maybe "NO_DSP_SIMD" is not even used in the code. https://github.com/f4exb/sdrangel/search?q=NO_DSP_SIMD&unscoped_q=NO_DSP_SIMD
Comment by Filipe Laíns (FFY00) - Saturday, 18 May 2019, 20:59 GMT
-DNO_DSP_SIMD=ON din't actually have any effect at all. -DENABLE_DISTRIBUTION=ON did limit SIMD to SSE2 and in some rare cases SSE3 which is fine by me.
Comment by Filipe Laíns (FFY00) - Saturday, 18 May 2019, 21:14 GMT
Ca you try 4.8.0-2? Thanks.
Comment by sven (commonuser) - Saturday, 18 May 2019, 21:52 GMT
Yes, works fine. Thank you :).
Comment by PyroPeter (pyropeter) - Sunday, 25 August 2019, 10:26 GMT
  • Field changed: Percent Complete (100% → 0%)
Regression in sdrangel 4.11.4-2 (fix for this bug seems to have vanished)
Comment by loqs (loqs) - Sunday, 25 August 2019, 12:28 GMT
@pyropeter if you add -DENABLE_GENERIC=ON to the cmake invokation does that work?
Comment by Filipe Laíns (FFY00) - Monday, 26 August 2019, 07:41 GMT
Yes, this was a regression. I must have messed up with svn conflicts.
Comment by Filipe Laíns (FFY00) - Monday, 26 August 2019, 07:45 GMT
Nop, `ENABLE_DISTRIBUTION` was removed by the upstream. I will need to look into it.
Comment by Filipe Laíns (FFY00) - Friday, 29 November 2019, 12:28 GMT
So upstream removed that option without documenting anything which required me to do some digging but it's finally ready. I pushed two sdrangel packages: sdrangel, without simd instructions, and sdrangel-sse4.1 with SSE 4.1 instructions.

Can you confirm sdrangel 4.12.1-1 resolves your problem?

Loading...