Segmentation fault

[guolas]

Hi,

I have found some weird behaviour in a code I am using right now. This code can be found in:

complete.jl

In the only file that there is in that repository, at the end of it, there are some parameters that define the execution, in particular this three:

number_of_bs = 10
antennas_tx = 4
antennas_rx = 2

If the parameter antennas_tx is changed to 3 and number_of_bs is set to 10, 11, or 12, executing this code will raise a Segmentation Fault.

Sometimes the error message is not a Segmentation Fault, but something like:

julia(41398,0x7fff70751180) malloc: *** error for object 0x10d807008: incorrect checksum for freed object - object was probably modified after being freed.
*** set a breakpoint in malloc_error_break to debug
Stack dump:
0.  Running pass 'X86 DAG->DAG Instruction Selection' on function '@julia_anonymous613'
Abort trap: 6

Or:

julia(41428,0x7fff70751180) malloc: *** error for object 0x10a7a2808: incorrect checksum for freed object - object was probably modified after being freed.
*** set a breakpoint in malloc_error_break to debug
Stack dump:
0.  Running pass 'Slot index numbering' on function '@julia_ctranspose'
Abort trap: 6

For other parameters I have tried, it works perfectly fine, and without any problem.

Regards,

Juanjo.

[pao]

For single files like this, you may find it easier to use a gist: https://gist.github.com/

[JeffBezanson]

@ViralBShah @andreasnoackjensen Could you take a look? This change fixed the segfault for me but I wonder if there are more cases of this, or if I did something wrong.

[andreasnoack]

I am on the road today so I cannot look into it before tomorrow. It is possible that the work space is not large enough, but there has also been weird errors in xgesdd. It calls xlasd4 which is buggy but I have not been able to make xgesdd fail consistently. Usually it had happened inside larger pieces of code.

[ViralBShah]

#2340 is the xlasd4 issue for which there is no resolution yet.