Limits and retention

These are deployment settings, so the numbers where you are running may differ. If an upload is rejected you are told which limit it hit, not just that it failed.

Some directories are never taken, because they are never useful to a run and one of them alone can exhaust the file count on a real repository:

.git   .hg   .svn   node_modules   __pycache__   .venv   venv
.mypy_cache   .pytest_cache   .tox   .idea   .next
build   dist   .cache

Beyond that, three kinds of entry are rejected outright rather than filtered:

• Paths that escape the archive. An entry pointing outside the folder it claims to be in is refused; it is a known way to write files where they do not belong.
• Symbolic links. They can point anywhere, so they are not extracted.
• Executable bits. Stripped from everything; nothing in an uploaded archive needs to be runnable.

Archives downloaded from a Git host arrive wrapped in a single folder named after the repo and branch, such as fifo_project-main/. That wrapper is stripped so your paths look the way they do locally.

It is not stripped when the single top-level directory is real project content. Names like rtl, src, tb, sim, docs and constraints are left alone; moving them up a level would break every relative include in your sources.

Retention is measured from real activity rather than from a bookkeeping timestamp, and the files are re-checked before anything is removed; a workspace you touched recently is left alone even if its record says otherwise.