Corrupted EDF file leads to segfault

[danjjl]

A corrupted EDF file causes python to segfault with the following error:

python(14094,0x1098325c0) malloc: *** error for object 0x7faae04a9e00: pointer being freed was not allocated
python(14094,0x1098325c0) malloc: *** set a breakpoint in malloc_error_break to debug
Abort trap: 6

edfbrowser reports the following error on the EDF file:

Datarecord duration is 1.000000000000 but timestep between
datarecord 70407 and preceding datarecord is 1.678000000000.
Offset from start of file: 0x47DC9E58

Stopped at the first error, additional errors may be present.

A fix to this bug would be to throw an exception instead of crashing python.

[skjerns]

do you have a sample file?

the error is thrown from the c-library (edflib). to catch it we will probably need to add some additional code in the edflib library.

[danjjl]

Here is a fake corrupted file that generates a maloc
test_segfault.edf.zip

(File is actually corrupted, i.e. file does not meet EDF+ specification --> suggestion is not to fix loading of this particular file but more not to segfault and cleanly report an error)

[skjerns]

As I suspected, the segfault cannot be caught in Python.

I would need to alter the C-source code of pyedflib and throw a Python exception there. I don't want to break anything by tempering with edflib, so at this moment it's one one of the priorities.

[DimitriPapadopoulos]

The issue should have been opened against upstream EDFlib. It looks like recent unit tests of that library do report an error without crashing when reading your file. Hopefully, this segfault will go away when upgrading to a more recent version 1.24 of that library (#230).