[Web] onnxruntime-node Linux addon binaries contain duplicate identical copies of libonnxruntime.so.x taking up extra ~40MB

[rotemdan]

Describe the issue

Linux (x64):

Linux (arm64):

I tested the files libonnxruntime.so.1 and libonnxruntime.so.1.21.0 are identical. On 1.21.0, SHA256 hashes:

x64:

05BFDDEEFC536A44478C20226218DAE8853995518BB2A994B4FA15A996F5F412

arm64:

36671B53885A8433882C4DB360138BD060F3580F7D7CB6578050E755E83B6128

This duplication seems to have started in 1.19.0. It's not there in 1.18.0

Can anything be done to reduce that extra ~40MB from the installed package size? Is it necessary?

I guess the gzip compression could with the npm (downloaded) package size, but I thought I should report this since I haven't seen it reported before (sorry if this is a duplicate).

To reproduce

See:

https://www.npmjs.com/package/onnxruntime-node/v/1.21.0?activeTab=code

Urgency

No response

ONNX Runtime Installation

Released Package

ONNX Runtime Version or Commit ID

1.21.0

Execution Provider

Other / Unknown

[rotemdan]

Alright, I've made an experiment to see the reduction in npm .tgz archive size with and without duplicates:

I've ran:

npm pack onnxruntime-node

Which resulted in a 78,144KB .tgz file.

Then I extracted it, removed the duplicates and ran:

tar -czvf x.tgz onnxruntime-node-1.21.0

Which resulted in a 62,674KB archive (62,512KB with maximum compression level).

So by removing the duplicates, the reduction in size in the compressed archive is about 15.4 MB, which is about 20% of original.