unix% aconvolve norm.fits acon_out.fits method=fft kernelspec="lib:gaus(2,5,1,10,10" # 30826: Received error signal SIGSEGV-segmentation violation. # 30826: An invalid memory reference was made. # 30826: segmentation fault: aconvolve (1) is: exit_upon_error->NULL
The size of the output convolved image may differ by a power of 2 between Unix and Linux when pad="yes".
When doing sliding-cell convolution (as opposed to fft), there is a bug that's triggered when the 1st element in the array is negative, e.g.
unix% aconvolve in out "txt:(-1,1)" meth=slide norm=none
For most settings of the edges parameter, the bug causes the output to be all NaNs. If edges is set to renorm, the bug will cause the output to be all zeros.
Other examples which trigger the bug include "(-5,3,2)" and "((-8,1,1),(1,1,1),(1,1,1))".