tipi-build / elfshaker Goto Github PK

Motivation

          👍 ❤️ 

Looking good simple and effective, I think though that we should use the file modification time metadata we have as an heuristic, because I have implemented similar things for the cache in the past using file checksum and it is impossible to do even the sha1sum of all source entries, it's very slow, that's why we rely on git.

The problem is not the sha1 computation is the IO heaviness, you basically read in RAM all built files and they are HUGE. I think this code could crash on typical Debug builds including a bit of boost libs.

If already the sha1sum should be computed on mmapped files or on streamed file operations.

If it didn't change it's the same (as elfshaker set file times on extract) and if it changed then we check the sha1. I haven't measured yet and we should before optimizing.

Originally posted by @daminetreg in #6 (review)

Description

• Measure with and without the optimization on big build trees (boost in Debug mode and so on )
• Implement mtime optimization file list diffiing
• Integrate in tipi

Definition Of Done

tipi can use the optimized elfshaker status