On a related note, my T1-weighted image is actually a 7-T MP2RAGE UNI image.
Is it still appropriate to run recon-all on a 7-T image?

From: fsbuild <fsbuild@contbay.com>
Sent: Tuesday, April 21, 2026 1:48
To: freesurfer@nmr.mgh.harvard.edu <freesurfer@nmr.mgh.harvard.edu>
Subject: [Freesurfer] Re: Recon-all error: free(): double free detected in tcache 2; Command terminated by signal 6

⚠ Caution: External sender

External Email - Use Caution

The "double free" notification may be from data corruption in memory, e.g., memory management in the Ubuntu Linux virtual machine running on top of the Windows host via WSL2 has gone wrong from trying to free the same location in memory 2X. So the Linux OS kills the process (signal 6). This might be from mri_synthseg running tensorflow and running out of memory. I would check on how much memory WSL2 has been given to run the guest Linux OS, Ubuntu 22. My guess is at least 4G of physical RAM needs to be reserved to run the Windows OS on the machine, and subsequently you could try allocating the rest to WSL2. I would try to give Ubuntu22 at least 16G of RAM, but for a full recon-all 32G ay be needed.

- R.

On Apr 20, 2026, at 11:14, Huaiyu Liu <huaiyu.liu@ucl.ac.uk> wrote: