This saves what, 2 KB out of a 600M release? Seems not worth it.

At some point we may want to extend that package function anyway. For example we use validate-release to check that our engine zips don't ship weird permissions that would be written on users' computers when unpacking our zips. So we may want to fix those permissions there once for all. On a side note not shipping the folders even remove the need to ship proper permissions on folders.

I also want to check if the macos engine zip isn't expected to ship symlinks (the macos deps file is already know to do that), preventing 7z to dereference them would save file space as well.

Right now when releasing the game I recompress all our dpk and zip files with my drt-zipbrute tool from there. It has an option to rewrite the permissions of files in zip (so they can pass through validate-release without failing), but it also offers many optimization techniques to reduce zip file size:

• it can touch all files in a zip to the same date: that doesn't improve file compression in zip itself but that improves zip file compression in zip, like engine/dpk/symbols zip getting better recompressed in universal zip,
• it can remove useless folder entries, not only it save file space by itself, it removes the need for having proper folder permissions, and of course the time unification a zip-in-zip optimization is just even not needed,
• it makes sure files are stored in the same order in the zip on every run (improves reproducibility),
• it can recompress separately each file with different tools (7z, zopfli…) and keep the smallest file for every file.

This not only make our zip valid for validate-release, it just produces the smallest zip possible.

I forgot to mention in the PR description that the sort call also makes sure the files are stored in the same order in the zip on every run, this improves reproducibility.