Version: 0.9.15
After a full extraction of a ~120-file corpus, graphify-out/manifest.json contained exactly 5 entries — precisely the 5 files that produced zero nodes — rather than the set of processed files.
Anything that reads the manifest for change detection or coverage accounting therefore sees a 5-file corpus, and a file that extracted successfully is indistinguishable from a file that was never seen.
Expected
Either the manifest lists every processed file along with its outcome, or the docs state plainly that it is a zero-node/failure report only — in which case the name manifest.json is misleading and something like zero-node-files.json would describe it better.
Version: 0.9.15
After a full extraction of a ~120-file corpus,
graphify-out/manifest.jsoncontained exactly 5 entries — precisely the 5 files that produced zero nodes — rather than the set of processed files.Anything that reads the manifest for change detection or coverage accounting therefore sees a 5-file corpus, and a file that extracted successfully is indistinguishable from a file that was never seen.
Expected
Either the manifest lists every processed file along with its outcome, or the docs state plainly that it is a zero-node/failure report only — in which case the name
manifest.jsonis misleading and something likezero-node-files.jsonwould describe it better.