When compiled with gcc (version 3.2.3 20030422) flag -fomit-frame-pointer "spice sample_circuit.cir" crashes with: internal error -- segmentation violation Maybe you should filter this gcc flag in your ebuild. Not sure if this is spice or gcc bug though. regards
Can you perhaps add some example file and give a step-by-step guide how to crash it?
On second thought this may also be some athlon/athlon mobo bug. Steps to reproduce it: step 1 Emerge spice with these flags in make.conf CFLAGS="-march=athlon-tbird -O3 -pipe -fomit-frame-pointer" step 2 Save example below as na.spc. ########na.spc############### network analysis example v1 1 0 dc 28 v2 3 0 dc 7 r1 1 2 5 r2 2 0 7 r3 2 3 6 .dc v1 28 28 1 .print dc v(1,2) v(2,0) v(2,3) .end ############################## Step 3 "spice -b na.spc" or spice na.spc output: Program: Spice, version: 3f5 Date built: Mon Nov 3 16:51:58 CET 2003 Type "help" for more information, "quit" to leave. Circuit: network analysis example internal error -- segmentation violation step 4: Emerge spice without -fomit-frame-pointer flag and its behaviour will be correct: Circuit: network analysis example Warning: can't parse '0': ignored Circuit: network analysis example Date: Mon Nov 3 17:14:14 2003 network analysis example DC transfer characteristic Mon Nov 3 17:14:14 2003 -------------------------------------------------------------------------------- Index sweep v(1)-v(2) v(2) v(2)-v(3) -------------------------------------------------------------------------------- 0 2.800000e+01 1.471963e+01 1.328037e+01 6.280374e+00 CPU time since last call: 0.020 seconds. Total CPU time: 0.020 seconds. Current dynamic memory usage = 139264, Dynamic memory limit = 4294967295. Any file with correct or incorrect syntax produces same error. I've heard before about -fomit-frame-pointer issues with some obscure software but this is my first problem with it.
The problem can be reproduced with CFLAGS="-march=athlon-xp -O3 -fomit-frame-pointer -pipe", but not with CFLAGS="-O3 -fomit-frame-pointer -pipe" (gcc 3.3.2). Optimization -O0 and -O1 don't crash. Looks like "athlon" and "omit-frame-pointer" is a bad combination for this package.
Since this does not depend on a single flag, but on a specific combination, I'd like to treat this as WONTFIX.