google-drive-ocamlfuse causes MS Word to create corrupt files

[astrada]

google-drive-ocamlfuse causes MS Word to create corrupt files

[astrada]

Sorry, I don't have Office, so I can't try it directly. Could you please do this test? Save the file on your filesystem, then copy it to the mounted directory. Is it still corrupted?

[pmorch]

Hi,

When I save it to the local directory, I make a note of its MD5 sum. I copy it to the mounted directory, and it keeps the MD5 sum. Both the original local and remote files can open fine.

If, instead, I save to the mounted directory, and then copy that file locally, both remote and local files have same MD5 sum and both are corrupt.

It seems it is the save process that causes the corruption. I have noticed that word does "odd" (but legal) tricks with saving to short-lived/temporary files in the same dir as the file during the save process.

Microsoft's Description of how Word creates temporary files e.g. says:

By saving to a temporary file first and then renaming the file to the proper name, Word ensures the data integrity of your original file against problems (such as a power failure or lost network connections) that may occur while the file is being written.

[pmorch]

I've also run it like so:

google-drive-ocamlfuse -debug -d <mountpoint> 2>&1 | tee google-drive-ocamlfuse.log

to give you detailed log on what is going on. Email me privately at [email protected], and I'll send you the log file, and a movie of what is going on in the GUI.

However, now it fails in another way. Instead of Word silently declaring the save a success, and then discovering corruption during open of the saved file, it now discovers the fail during save, asks for a new file name to save under, but still fails again. After that, Word gives up saving.

For some reason, after a few tries, I cannot reproduce my first scenario, but only this new one. Either way, though, Word is not able to save to a google-drive-ocamlfuse mountpoint.

[pmorch]

I can confirm that after commit 9d11a75 (in version 0.4.4), this issue has been resolved for me. Thanks, Alessandro