octave-3.2.3 crashes in eigs (arpack)

[jarno.rajahalme@…]

On OSX 10.6.3 (and likely earlier), running a 64-bit kernel, have not tested with 32-bit.

Octave 3.2.3 causes a segmentation fault when complex numbers are used with eigs(). within octave, type

octave:1> test eigs panic: Segmentation fault -- stopping myself... Segmentation fault $

The reason for this is that the macports arpack package, that octave depends on, is compiled with options incompatible with the octave, and other fortran compiled dependencies of octave. Either the dependency should be removed (as it is useless as is), or arpack needs to be updated.

[jarno.rajahalme@…]

Cc Me!

[jmroot (Joshua Root)]

Please remember to cc the maintainer, and note you do not need to be in cc when you are the reporter.

[jarno.rajahalme@…]

Replying to jmr@…:

Please remember to cc the maintainer, and note you do not need to be in cc when you are the reporter.

Just checked this again with octave 3.2.3_2, it says the function eigs() not included, and no crash. Still, arpack is compiled with incompatible options (-ff2c), and thus does not work as intended. When compiling octave 3.2.3 from sources, it checks for arpack, finds it, and crashes in test eigs.

[adfernandes (Andrew Fernandes)]

Cc Me!

[adfernandes (Andrew Fernandes)]

FYI - I rebuilt arpack using a fresh pull of macports (to fix #25186), and had no problems with octave's eigs command. My build is octave @3.2.4_0+gcc44. Mac OS 10.6.3, 64-bit build as per default.

[michaelld (Michael Dickens)]

If this is still really an issue ... please do:

sudo port selfupdate

and then try installing it again. Should be fixed with r114034.