MNE appends empty data at the end of Raw object after CNT importation

[withmywoessner]

Description of the problem

Hello! My lab has some old neuroscan cnt files that we are processing and I discovered a bug where MNE appends empty data to the end of the file after importation using mne.io.read_raw_cnt()

When I read the file in neuroscan the file is normal so I know it is not a problem with the file.

Steps to reproduce

import mne
mne.io.read_raw_cnt('empty_end.cnt', data_format = 'int32').plot()

Link to data

I cannot shorten the file because it removes the problem.
Here is a link to the file:
https://wustl.box.com/s/nfcbstcegtbh954ixh8orzyteiz0wxk2

Expected results

The expected results should be the data imports as normal and looks like this at the end:

Actual results

Instead there is empty data at the end and when I run:
mne.io.read_raw_cnt('/home/woess/current_assignment/empty_end.cnt', data_format = 'int32').to_data_frame().tail(100)
The data looks like this:

The plot looks like this:

Additional information

Platform Linux-5.15.90.2-microsoft-standard-WSL2-x86_64-with-glibc2.31
Python 3.10.10 | packaged by conda-forge | (main, Mar 24 2023, 20:08:06) [GCC 11.3.0]
Executable /home/woess/mambaforge/envs/mne/bin/python
CPU x86_64 (12 cores)
Memory 11.3 GB

Core
├☑ mne 1.4.0
├☑ numpy 1.23.5 (OpenBLAS 0.3.21 with 12 threads)
├☑ scipy 1.10.1
├☑ matplotlib 3.7.1 (backend=module://matplotlib_inline.backend_inline)
├☑ pooch 1.7.0
└☑ jinja2 3.1.2

Numerical (optional)
├☑ sklearn 1.2.2
├☑ numba 0.56.4
├☑ nibabel 5.0.1
├☑ nilearn 0.10.0
├☑ dipy 1.6.0
├☑ openmeeg 2.5.6
├☑ pandas 2.0.3
└☐ unavailable cupy

Visualization (optional)
├☑ pyvista 0.38.5 (OpenGL 3.3 (Core Profile) Mesa 21.2.6 via D3D12 (AMD Radeon(TM) Graphics))
├☑ pyvistaqt 0.9.1
├☑ ipyvtklink 0.2.2
├☑ vtk 9.2.6
├☑ qtpy 2.3.0 (PyQt5=5.15.6)
├☑ ipympl 0.9.3
├☑ pyqtgraph 0.13.2
└☑ mne-qt-browser 0.4.0

Ecosystem (optional)
└☐ unavailable mne-bids, mne-nirs, mne-features, mne-connectivity, mne-icalabel, mne-bids-pipeline

[withmywoessner]

I think a solution to this problem would be to change this cnt.py file to stop reading after n_samples samples: link to code
It is commented that this variable is not reliable but I am not sure why. Currently the file stops being read after sample_stop samples: link
Where sample stop is equal to sample_start + min(block_size // f_channels, data_left // f_channels - sample_start)
I wrote a short script to return the n_samples variable myself and the number of samples it returned exactly matched where the end of the file was supposed to be, so maybe this problem can be fixed by:
sample stop = sample_start + min((n_samples, block_size // f_channels, data_left // f_channels - sample_start))

[larsoner]

@withmywoessner want to try making this change and see if it breaks any tests?

We should probably add some kwarg to control whether the variable should be used or not... if someone wrote that it's unreliable then they probably have files where it's wrong.