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Do NOT report bugs when a package is just outdated, or it is in the AUR. Use the 'flag out of date' link on the package page, or the Mailing List.
REPEAT: Do NOT report bugs for outdated packages!
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Do NOT report bugs when a package is just outdated, or it is in the AUR. Use the 'flag out of date' link on the package page, or the Mailing List.
REPEAT: Do NOT report bugs for outdated packages!
FS#22314 - [erlang] R14B triggers "Illegal instruction" on pre-P4 x86
Attached to Project:
Arch Linux
Opened by Famo Qua (famoqua) - Tuesday, 04 January 2011, 18:34 GMT
Last edited by Vesa Kaihlavirta (vegai) - Monday, 14 March 2011, 10:01 GMT
Opened by Famo Qua (famoqua) - Tuesday, 04 January 2011, 18:34 GMT
Last edited by Vesa Kaihlavirta (vegai) - Monday, 14 March 2011, 10:01 GMT
|
DetailsDescription:
Default erlang build won't execute correctly on older processors than Pentium 4 and may raise "Illegal instruction" signals. Tested on Athlon XP 2500+ (without SSE2). Basically, configure script keeps unsupported instructions enabled by default. Additional info: extra/erlang R14B-1 Likely to affect earlier builds too. Steps to reproduce: Sadly I'm not into erlang programming language, but rabbitmq dies at startup with an "Illegal instruction" message. Proposed solution: Explicitly add: "--enable-ethread-pre-pentium4-compatibility enable_ethread_pre_pentium4_compatibilit=yes" to "./configure" before compiling "otp_src_R41B". Seems to be a known issue: http://erlang.2086793.n4.nabble.com/R14B-Illegal-instruction-td2544273.html A patched PKGBUILD definitely worked for my Athlon XP. |
This task depends upon
"--enable-ethread-pre-pentium4-compatibility enable_ethread_pre_pentium4_compatibilit=yes"
without the last 'y' in "compatibility" and without the double dash?
I did build this and will upload -2, it just seems odd.
The configure script should enable/disable the option accordingly but for some reason it doesn't (I'm not an autoconf expert). My comment is about "downgrading" optimizations to make the package work on a wider range of CPUs by default.