OneDrive on mac is keeping two copies of all downloaded files

Environment:

MacOS: 12.4 (21F79)

Mac filesystem type hosting OneDrive folder: APFS Data Volume: /dev/disk1s1 on /System/Volumes/Data (apfs, local, journaled, nobrowse)
OneDrive: 22.131.0619.0001 (Standalone)
OneDrive home account as part of Microsoft 365 Personal subscription
OneDrive folder: ~/OneDrive, symlink to /Users/mikaelkjaerulff/Library/CloudStorage/OneDrive-Personal
OneDrive “Files On-Demand” is off by selecting “Download all OneDrive files now”
OneDrive “File collaboration” is off
OneDrive status is “Your files are synced” and there are no warnings/errors
OneDrive storage: 134 GB used of 1.01 TB (as reported by https://onedrive.live.com/?v=managestorage)

I have discovered that OneDrive keeps 2 local copies of all the synced files.  So instead of taking up 134 GB of disk space it uses 268 GB – that can’t be right!  The (I assume) primary copy is in the OneDrive folder:

~/Library/CloudStorage/OneDrive-Personal/

And then there is the (I assume) secondary (or internal?) copy in:

~/Library/Group Containers/UBF8T346G9.OneDriveStandaloneSuite/OneDrive.noindex/OneDrive/

This is true copies of the files, not hard-linked files – I checked that it is different inodes:

> cd ~/Library

> ls -i CloudStorage/OneDrive-Personal/Gravhund/ActivitiesData.xlsx Group\ Containers/UBF8T346G9.OneDriveStandaloneSuite/OneDrive.noindex/OneDrive/Gravhund/ActivitiesData.xlsx
8965664803 CloudStorage/OneDrive-Personal/Gravhund/ActivitiesData.xlsx
8965664807 Group Containers/UBF8T346G9.OneDriveStandaloneSuite/OneDrive.noindex/OneDrive/Gravhund/ActivitiesData.xlsx

There are some differences. If I run a diff on the two locations, like:

> cd ~/Library

> diff -rq CloudStorage/OneDrive-Personal/ Group\ Containers/UBF8T346G9.OneDriveStandaloneSuite/OneDrive.noindex/OneDrive | grep -v .DS_Store > ~/tmp/diff.txt

Note that I exclude Mac Spotlight files (.DS_Store) – Spotlight does not run on the secondary copy. I get:

> cat -v ~/tmp/diff.txt

Only in CloudStorage/OneDrive-Personal/: .Trash

Only in Group Containers/UBF8T346G9.OneDriveStandaloneSuite/OneDrive.noindex/OneDrive/Ejerforeningen matr. nr. 146 Hareskov By, Værløse/Billag/2021: 2021-01-14 ?M-^Ersoversigt.PDF
Only in CloudStorage/OneDrive-Personal/Ejerforeningen matr. nr. 146 Hareskov By, Værløse/Billag/2021: 2021-01-14 A?M-^Jrsoversigt.PDF
Only in Group Containers/UBF8T346G9.OneDriveStandaloneSuite/OneDrive.noindex/OneDrive/Ejerforeningen matr. nr. 146 Hareskov By, Værløse/Billag/2021: 2021-02-17 SEAS-NVE - årsopgørelse.pdf
Only in CloudStorage/OneDrive-Personal/Ejerforeningen matr. nr. 146 Hareskov By, Værløse/Billag/2021: 2021-02-17 SEAS-NVE - a?M-^Jrsopgørelse.pdf
Only in Group Containers/UBF8T346G9.OneDriveStandaloneSuite/OneDrive.noindex/OneDrive/Ejerforeningen matr. nr. 146 Hareskov By, Værløse/Billag/2021: 2021-03-22 Novafos - årsregning 2020.pdf
Only in CloudStorage/OneDrive-Personal/Ejerforeningen matr. nr. 146 Hareskov By, Værløse/Billag/2021: 2021-03-22 Novafos - a?M-^Jrsregning 2020.pdf
Only in Group Containers/UBF8T346G9.OneDriveStandaloneSuite/OneDrive.noindex/OneDrive/Ejerforeningen matr. nr. 146 Hareskov By, Værløse/Billag/2021: 2021-12-31 ?M-^Ersoversigt.PDF
Only in CloudStorage/OneDrive-Personal/Ejerforeningen matr. nr. 146 Hareskov By, Værløse/Billag/2021: 2021-12-31 A?M-^Jrsoversigt.PDF
Files CloudStorage/OneDrive-Personal/Gravhund/GravDB/ETL/IMPO/sed0.txt and Group Containers/UBF8T346G9.OneDriveStandaloneSuite/OneDrive.noindex/OneDrive/Gravhund/GravDB/ETL/IMPO/sed0.txt differ
Files CloudStorage/OneDrive-Personal/Gravhund/GravDB/ETL/IMPO/sed2.txt and Group Containers/UBF8T346G9.OneDriveStandaloneSuite/OneDrive.noindex/OneDrive/Gravhund/GravDB/ETL/IMPO/sed2.txt differ
Files CloudStorage/OneDrive-Personal/Gravhund/GravDB/ETL/KULD/sed0.txt and Group Containers/UBF8T346G9.OneDriveStandaloneSuite/OneDrive.noindex/OneDrive/Gravhund/GravDB/ETL/KULD/sed0.txt differ
Files CloudStorage/OneDrive-Personal/Gravhund/GravDB/ETL/KULD/sed2.txt and Group Containers/UBF8T346G9.OneDriveStandaloneSuite/OneDrive.noindex/OneDrive/Gravhund/GravDB/ETL/KULD/sed2.txt differ
Files CloudStorage/OneDrive-Personal/Gravhund/GravDB/ETL/PROV/sed2.txt and Group Containers/UBF8T346G9.OneDriveStandaloneSuite/OneDrive.noindex/OneDrive/Gravhund/GravDB/ETL/PROV/sed2.txt differ
Files CloudStorage/OneDrive-Personal/Gravhund/GravDB/ETL/PROV.KRITIK/sed0.txt and Group Containers/UBF8T346G9.OneDriveStandaloneSuite/OneDrive.noindex/OneDrive/Gravhund/GravDB/ETL/PROV.KRITIK/sed0.txt differ
Files CloudStorage/OneDrive-Personal/Gravhund/GravDB/ETL/PROV.PRID/sed2.txt and Group Containers/UBF8T346G9.OneDriveStandaloneSuite/OneDrive.noindex/OneDrive/Gravhund/GravDB/ETL/PROV.PRID/sed2.txt differ
Files CloudStorage/OneDrive-Personal/Gravhund/GravDB/ETL/RYGI/sed0.txt and Group Containers/UBF8T346G9.OneDriveStandaloneSuite/OneDrive.noindex/OneDrive/Gravhund/GravDB/ETL/RYGI/sed0.txt differ
Files CloudStorage/OneDrive-Personal/Gravhund/GravDB/ETL/RYGI/sed2.txt and Group Containers/UBF8T346G9.OneDriveStandaloneSuite/OneDrive.noindex/OneDrive/Gravhund/GravDB/ETL/RYGI/sed2.txt differ
Files CloudStorage/OneDrive-Personal/Gravhund/GravDB/ETL/RYGS/sed2.txt and Group Containers/UBF8T346G9.OneDriveStandaloneSuite/OneDrive.noindex/OneDrive/Gravhund/GravDB/ETL/RYGS/sed2.txt differ
Files CloudStorage/OneDrive-Personal/Gravhund/GravDB/ETL/UDST/sed1.txt and Group Containers/UBF8T346G9.OneDriveStandaloneSuite/OneDrive.noindex/OneDrive/Gravhund/GravDB/ETL/UDST/sed1.txt differ
Files CloudStorage/OneDrive-Personal/Gravhund/GravDB/ETL/UDST/sed2.txt and Group Containers/UBF8T346G9.OneDriveStandaloneSuite/OneDrive.noindex/OneDrive/Gravhund/GravDB/ETL/UDST/sed2.txt differ
Only in CloudStorage/OneDrive-Personal/Gravhund/WWW/nsdgk: index.htm~
Only in CloudStorage/OneDrive-Personal/Gravhund/WWW/nsdgk: tilmeld.htm~
Only in CloudStorage/OneDrive-Personal/Gravhund/WWW/nsdgk: tilmeld_ver.htm~
Only in CloudStorage/OneDrive-Personal/Gravhund/WWW/nsdgk/traen: lydighed.htm~
Only in CloudStorage/OneDrive-Personal/Gravhund/WWW/nsdgk/traen: ring.htm~
Only in CloudStorage/OneDrive-Personal/Gravhund/WWW/nsdgk/traen: spor.htm~

I see the following main differences:

1. Symlinks in primary are converted into normal files with content of the symlink in the secondary.
2. Some file names are converted (like ?M-^Ersoversigt.PDF to årsregning 2020.pdf).
3. Looks like some emacs backup-files are not copied to the secondary location and not uploaded to the cloud.

(1)

 Symlinks are converted to normal files:

> ls -l "CloudStorage/OneDrive-Personal/Gravhund/GravDB/ETL/IMPO/sed0.txt" "Group Containers/UBF8T346G9.OneDriveStandaloneSuite/OneDrive.noindex/OneDrive/Gravhund/GravDB/ETL/IMPO/sed0.txt"
lrwxrwxr-x  1 mikaelkjaerulff  staff  18 Jun 17 20:01 CloudStorage/OneDrive-Personal/Gravhund/GravDB/ETL/IMPO/sed0.txt -> ../common/sed0.txt
-rwxr-xr-x@ 1 mikaelkjaerulff  staff  18 Jun 17 20:01 Group Containers/UBF8T346G9.OneDriveStandaloneSuite/OneDrive.noindex/OneDrive/Gravhund/GravDB/ETL/IMPO/sed0.txt
> cat -v "Group Containers/UBF8T346G9.OneDriveStandaloneSuite/OneDrive.noindex/OneDrive/Gravhund/GravDB/ETL/IMPO/sed0.txt"
../common/sed0.txt

I do understand that it often makes little to no sense synchronizing symbolic links to a remote location as-is – they will often point to an invalid location.  But if the symlink is relative (like above), then it might work (if the remote filesystem supports the concept of symbolic links).  The real annoyance is that if you do a fresh download from the OneDrive cloud the symlinks are “lost” – which is the case when you do a reset of OneDrive.  The symlinks are restored as normal files with the content of the symlink.  As far as I can tell there is no easy way to find just these files and convert them back to symlinks.

(2)

Apple’s APFS file system encodes file and directory names using Unicode but is not a “normalizing file system”. I assume the secondary OneDrive copy does some normalization (from NFC to NFD) of file and directory names before uploading to the cloud.  At least that is the only explanation I can come up with.

Ref on APFS and normal forms:

• https://en.wikipedia.org/wiki/Unicode\_equivalence#Normal\_forms
• https://eclecticlight.co/2021/05/08/explainer-unicode-normalization-and-apfs/
• https://gist.github.com/JamesChevalier/8448512
• https://medium.com/@sthadewald/the-utf-8-hell-of-mac-osx-feef5ea42407

(3)

I don’t understand why emacs backup files are not copied to the secondary and subsequently uploaded to the cloud.  I have a suspicion it could be due to the way emacs create the backup file.  Nevertheless, OneDrive does not pick up the emacs backup files – this must be a bug in OneDrive.

I could understand if OneDrive has the need for a temporary location (to normalize for example) for files and directories before being uploaded to the cloud – but with emphasis on “temporary”.  The secondary copies live forever – that can’t be right.

As a side note: Still OneDrive goes to 100+ % CPU for a minute or so every time I save a file in my OneDrive location.  At least the newest version of OneDrive does not cause kernel panics anymore (fingers crossed, knock on wood…).

What – if anything – can I do to avoid having the secondary copies?

/Mikael