Always canonicalize glob paths passed to Tera #799
Merged
Conversation
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
This is a regression test for Keats#574.
|
The CI complains that |
|
Ah it's because the CI still uses rustc version 1.56.0. I can change the |
|
Yep, no need to bump the MSRV for such a small change. I remember some issues with canonicalizing paths on Windows but my memory is fuzzy :/ |
This is to work around an issue in globwalk (Gilnaa/globwalk#28) due to which paths starting with `./` or `../` never match anything. This fixes Keats#574.
eguiraud
force-pushed
the
always-canonicalize-globs
branch
from
27e5a28 to
7bb748a
Compare
|
Changed the |
Keats
approved these changes
Feb 22, 2023
|
Hi, is this ready to merge? 👀 |
|
Yes, i just needed to block some time to merge all the PR and do a release |
vimpostor
added a commit
to vimpostor/tera
that referenced
this pull request
Apr 6, 2023
Regression was introduced in 9479c28. Since that commit providing an invalid glob to tera will make it fail to create an instance, first (in that commit) by making it panic and since 1f95878 by making it return an error. While returning an error is not entirely unlogical, it doesn't match what most languages do with invalid globs. - Bash will by default return an empty set on invalid globs, as the `failglob` option is off by default - Python will likewise return the empty set instead of throwing an exception, when doing something like `glob.glob("/dev/null/*")` - Rust's `globwalk` will also not error, but return an empty set In fact, we use globwalk in tera and the only reason we panic is by accident, because `std::fs::canonicalize()` checks the path. It is better to match other language's glob behaviour, therefore we resort back to the original path if `canonicalize()` fails to return the empty set again when encountering an invalid glob. Even more so, since this caused a lot of regressions already, including in zola. Fixes getzola/zola#2150 ref: Keats#819 ref: Keats#820 ref: Keats#799
vimpostor
added a commit
to vimpostor/tera
that referenced
this pull request
Apr 6, 2023
Regression was introduced in 9479c28. Since that commit, providing an invalid glob to tera will make it fail to create an instance, first (in that commit) by making it panic and since 1f95878 by making it return an error. While returning an error is not entirely unlogical, it doesn't match what most languages do with invalid globs. - Bash will by default return an empty set on invalid globs, as the `failglob` option is off by default - Python will likewise return the empty set instead of throwing an exception, when doing something like `glob.glob("/dev/null/*")` - Rust's `globwalk` will also not error, but return an empty set In fact, we use globwalk in tera and the only reason we panic is by accident, because `std::fs::canonicalize()` checks the path. It is better to match other language's glob behaviour, therefore we resort back to the original path if `canonicalize()` fails to return the empty set again when encountering an invalid glob. We should especially restore the previous behaviour, as this caused a lot of regressions already, including in zola. Fixes getzola/zola#2150 ref: Keats#819 ref: Keats#820 ref: Keats#799
vimpostor
added a commit
to vimpostor/tera
that referenced
this pull request
Apr 6, 2023
Regression was introduced in 9479c28. Since that commit, providing an invalid glob to tera will make it fail to create an instance, first (in that commit) by making it panic and since 1f95878 by making it return an error. While returning an error is not entirely unlogical, it doesn't match what most languages do with invalid globs. - Bash will by default return an empty set on invalid globs, as the `failglob` option is off by default - Python will likewise return the empty set instead of throwing an exception, when doing something like `glob.glob("/dev/null/*")` - Rust's `globwalk` will also not error, but return an empty set In fact, we use globwalk in tera and the only reason we panic is by accident, because `std::fs::canonicalize()` checks the path. It is better to match other language's glob behaviour, therefore we resort back to the original path if `canonicalize()` fails. We should especially restore the previous behaviour, as this caused a lot of regressions already, including in zola. Fixes getzola/zola#2150 ref: Keats#819 ref: Keats#820 ref: Keats#799
vimpostor
added a commit
to vimpostor/tera
that referenced
this pull request
Apr 10, 2023
Regression was introduced in 9479c28. Since that commit, providing an invalid glob to tera will make it fail to create an instance, first (in that commit) by making it panic and since 1f95878 by making it return an error. While returning an error is not entirely unlogical, it doesn't match what most languages do with invalid globs. - Bash will by default return an empty set on invalid globs, as the `failglob` option is off by default - Python will likewise return the empty set instead of throwing an exception, when doing something like `glob.glob("/dev/null/*")` - Rust's `globwalk` will also not error, but return an empty set In fact, we use globwalk in tera and the only reason we panic is by accident, because `std::fs::canonicalize()` checks the path. It is better to match other language's glob behaviour, therefore we resort back to the original path if `canonicalize()` fails. We should especially restore the previous behaviour, as this caused a lot of regressions already, including in zola. Fixes getzola/zola#2150 ref: Keats#819 ref: Keats#820 ref: Keats#799
Keats
pushed a commit
that referenced
this pull request
Apr 16, 2023
Regression was introduced in 9479c28. Since that commit, providing an invalid glob to tera will make it fail to create an instance, first (in that commit) by making it panic and since 1f95878 by making it return an error. While returning an error is not entirely unlogical, it doesn't match what most languages do with invalid globs. - Bash will by default return an empty set on invalid globs, as the `failglob` option is off by default - Python will likewise return the empty set instead of throwing an exception, when doing something like `glob.glob("/dev/null/*")` - Rust's `globwalk` will also not error, but return an empty set In fact, we use globwalk in tera and the only reason we panic is by accident, because `std::fs::canonicalize()` checks the path. It is better to match other language's glob behaviour, therefore we resort back to the original path if `canonicalize()` fails. We should especially restore the previous behaviour, as this caused a lot of regressions already, including in zola. Fixes getzola/zola#2150 ref: #819 ref: #820 ref: #799
theapplegates
pushed a commit
to theapplegates/sequoia
that referenced
this pull request
Mar 27, 2025
Installing sq from crates.io (cargo install sequoia-sq) was broken by a semver-compatible change in Tera. Running cargo test uses the lockfile and isn't affected. This has the side benefit of reducing dependency bloat, the baseline depends on check/build/build --release but in the case of a non-release build the dependency count goes from 403 to 315. Fixes #2. The subplot/tera issue was likely triggered by this change in tera 1.18: <Keats/tera#799>.
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
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.
This is to work around an issue in globwalk (Gilnaa/globwalk#28) due to which paths starting with
./or../never match anything. A regression test has been added as well.This fixes #574.