Bug: Relative template_dir path causes TemplateNotFound error in monorepo configuration #845
Description
Activity
@deme3 Thanks for the bug report and the initial troubleshooting you have accomplished. I have been working an extensive rewrite of the changelog testing and this will definitely be one of the new tests to add along with a fix.
Reacted by JonZeolla@codejedi365 thanks for letting me know. :)
Until we get the new release with the changelog rewrite, I can suggest the following patch for whoever is using this tool in a monorepo like me (note that I have no idea if this is good, I haven't looked much into the codebase, I just needed a quick fix for now):
unchanged: --- a/cli/commands/changelog.py +++ b/cli/commands/changelog.py @@ -3,6 +3,7 @@ from __future__ import annotations import logging import os from typing import TYPE_CHECKING +from pathlib import Path import click @@ -76,7 +77,8 @@ def changelog(ctx: click.Context, release_tag: str | None = None) -> None: f"{template_dir!r} relative to {repo.working_dir!r}" ) else: - recursive_render(template_dir, environment=env, _root_dir=repo.working_dir) + project_root = Path(runtime.global_cli_options.config_file).absolute().parent + recursive_render(template_dir, environment=env, _root_dir=project_root) if release_tag: if runtime.global_cli_options.noop: @@ -116,4 +118,4 @@ def changelog(ctx: click.Context, release_tag: str | None = None) -> None: ) except Exception as e: log.exception(e) - ctx.fail(str(e)) \ No newline at end of file + ctx.fail(str(e)) unchanged: --- a/cli/config.py +++ b/cli/config.py @@ -395,9 +395,10 @@ class RuntimeContext: ) # changelog_file + project_root = Path(global_cli_options.config_file).absolute().parent changelog_file = Path(raw.changelog.changelog_file).resolve() - template_dir = Path(repo.working_tree_dir or ".") / raw.changelog.template_dir + template_dir = Path(project_root or ".") / raw.changelog.template_dir template_environment = environment( template_dir=raw.changelog.template_dir, @@ -434,4 +435,4 @@ class RuntimeContext: # credential masker self.apply_log_masking(self.masker) - return self \ No newline at end of file + return self only in patch2: unchanged: --- a/cli/commands/version.py +++ b/cli/commands/version.py @@ -5,6 +5,7 @@ import os import subprocess from contextlib import nullcontext from datetime import datetime +from pathlib import Path from typing import TYPE_CHECKING, ContextManager, Iterable import click @@ -402,6 +403,7 @@ def version( # noqa: C901 hvcs_client=hvcs_client, release_history=rh ) changelog_context.bind_to_environment(env) + project_root = Path(runtime.global_cli_options.config_file).absolute().parent updated_paths: list[str] = [] if update_changelog: @@ -409,13 +411,13 @@ def version( # noqa: C901 if opts.noop: noop_report( f"would have recursively rendered the template directory " - f"{template_dir!r} relative to {repo.working_dir!r}. " + f"{template_dir!r} relative to {project_root!r}. " "Paths which would be modified by this operation cannot be " "determined in no-op mode." ) else: updated_paths = recursive_render( - template_dir, environment=env, _root_dir=repo.working_dir + template_dir, environment=env, _root_dir=project_root ) else: @@ -425,13 +427,13 @@ def version( # noqa: C901 if opts.noop: noop_report( "would have written your changelog to " - + str(changelog_file.relative_to(repo.working_dir)) + + str(changelog_file.relative_to(project_root)) ) else: changelog_text = render_default_changelog_file(env) changelog_file.write_text(changelog_text, encoding="utf-8") - updated_paths = [str(changelog_file.relative_to(repo.working_dir))] + updated_paths = [str(changelog_file.relative_to(project_root))] if commit_changes and opts.noop: noop_report(
which instead of using the repository root directory as reference, uses the folder where the config is located.
Reacted by RoFzReacted by codejedi365 and JonZeolla@deme3 I hope to have this in next week. My PR has been pending since the weekend.
I was a bit concerned that this could break other changelog designs but we didn't have any testing in place to validate either scenario
Reacted by RoFz and Demetrio BattagliaIt has been 60 days since the last update on this confirmed issue. @python-semantic-release/team can you provide an update on the status of this issue?
Need to circle back around to this issue. It slipped my mind, unfortunately the patch above will not work anymore given the refactor I did in v9.8.4 but if you port it to the changelog_writer.py file it should be easy enough to get back to your fix. I don't have a good pythonic monorepo example to go off of which is been part of the delay in support of monorepos. If you have an example open source project I can mirror for testing that would be great.
Reacted by JonZeolla@codejedi365 https://github.com/langchain-ai/langchain/tree/master/libs is a pretty good open source python monorepo
@JonZeolla, thank you!
It has been 60 days since the last update on this confirmed issue. @python-semantic-release/team can you provide an update on the status of this issue?
Still in backlog
5 remaining items
It has been 60 days since the last update on this confirmed issue. @python-semantic-release/team can you provide an update on the status of this issue?
Still in backlog
It has been 60 days since the last update on this confirmed issue. @python-semantic-release/team can you provide an update on the status of this issue?
- changed the title [-]Relative template_dir path causes TemplateNotFound error in monorepo configuration[/-] [+]Bug: Relative template_dir path causes TemplateNotFound error in monorepo configuration[/+] on Jun 18, 2025
Still in backlog
It has been 90 days since the last update on this confirmed issue. @python-semantic-release/team can you provide an update on the status of this issue?
Still in the backlog
It has been 90 days since the last update on this confirmed issue. @python-semantic-release/team can you provide an update on the status of this issue?
The problem
TemplateNotFound error thrown by Jinja2 when generating changelog in a subfolder. The directory structure is as follows:
- main_program +-- templates +-- main_program +-- CHANGELOG.md.j2 +-- CHANGELOG.md +-- pyproject.toml +-- ... - lib1 - lib2To be clear, the templates are discovered by semantic-release, but when they're passed to Jinja, a relative path is used, and Jinja is not able to find them.
Expected behavior
I expect a generated CHANGELOG.md in the same level as pyproject.toml by using the templates provided in the configuration.
Environment
- Python Version: v3.8.18
- Pip Version: v23.0.1
- Semantic Release Version: v9.1.0
pip freezeUsing build==1.0.3 and setuptools=56.0.0.
Configuration
Logs
GitHub Attachment to Logs
Additional context
There's something wrong here:
python-semantic-release/semantic_release/changelog/template.py
Lines 77 to 123 in 82bfcd3
specifically, line 109
python-semantic-release/semantic_release/changelog/template.py
Lines 107 to 110 in 82bfcd3