Summary: | import-w3c-tests should work even with broken symlinks in WPT repo | ||
---|---|---|---|
Product: | WebKit | Reporter: | Tim Nguyen (:ntim) <ntim> |
Component: | Tools / Tests | Assignee: | Nobody <webkit-unassigned> |
Status: | NEW --- | ||
Severity: | Normal | CC: | ap, gsnedders, webkit-bug-importer |
Priority: | P5 | Keywords: | InRadar |
Version: | WebKit Nightly Build | ||
Hardware: | Unspecified | ||
OS: | Unspecified | ||
Bug Depends on: | 237724 | ||
Bug Blocks: |
Description
Tim Nguyen (:ntim)
2022-12-12 07:59:46 PST
This is frankly bizarre/worrying, as I don't see why we'd be trying to copy a file that doesn't exist. As I commented in the upstream PR of the revert (https://github.com/web-platform-tests/wpt/pull/37442): > [ef70ce9](https://github.com/web-platform-tests/wpt/commit/ef70ce9623f50e9e7a89bb713920588cfb1259cf) left `tools/third_party/funcsigs/docs/index.rst` as a symlink to the (non-existent, removed) `tools/third_party/funcsigs/docs/README.rst`, which I believe is the origin of all the problems here. Given WPT _shouldn't_ have any broken symlinks, this is very low priority, and we're better off fixing this in the near-term by just reverting any change to WPT that introduces and broken symlinks. The importer bug here would be easy to fix with the Python 3.3+ follow_symlinks argument on `shutil.copyfile(src, dst, *, follow_symlinks=True)`, as we just need to pass `follow_symlinks=False` in copy_tests in test_downloader.py (and pass it though webkitpy.common.system.filesystem), but as long as we support Python 2 this is more effort than is probably justified. For the record, https://github.com/web-platform-tests/wpt/pull/41292 re-lands the funcsigs drop _with_ the symlink actually removed this time. |