Don't canonicalise file keys#1250
Open
lionel- wants to merge 1 commit into
Open
Conversation
This was referenced Jun 3, 2026
Only canonicalise when comparing against paths from external sources (e.g. normalised by R). Canonicalisation may corrupt the identity of paths sent by frontends, and may prevent matching files at all when they are deleted (at which point a non-canonicalised path sent by the frontend can't be matched against canonicalised paths we've stored as keys)
lionel-
force-pushed
the
oak/salsa-14-canonicalisation
branch
from
0e6a188 to
eb2c844
Compare
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Branched from #1245
Progress towards #1212
While I was working on breakpoints I ran into difficulties matching files from different locations: frontend URIs via DAP requests and source references from R whose paths went through
normalizePath(). The latter does file canonicalisation, which means that in addition to resolving.and.., it also follows symlinks. The frontend URIs are not canonicalised, which caused CI failures when running tests in projects storedc in temporary folders, with mismatches between/tmpand/private/tmp. The fix I implemented was to canonicalise everywhere and encode that canonicalisation happened with theUrlIdfile, which we stored as file keys in our maps tracking state (e.g. breakpoints) by file.The
UrlIdnormalisation also dealt with other issues that surfaced over time, such as windows letter drive casing because to fix issues with lowercase on one end and uppercase on the other.The same sort of issue came up with the LSP: we get URIs from the frontend, but we also encounter
source()calls that may be relative and we need to resolve them. So I reached for the sameUrlIdapproach to make the different file paths agree with each other.It turns out that doing any sort of normalisation that requires I/O is a red flag. If a file gets deleted on disk, and we get an event about that file, there is literally no way to match it back to what we know. For the LSP that's a real issue because we get file deletion events after the file is deleted.
The other thing to watch out for: Always report back the unormalised URI/path to the frontend, because that's what they use to map files on their end. Any changes might cause the frontend to treat the file as different.
With this PR:
.and... That matches r-a and ty.