* Qual: Adjust cache configuration for windows-ci
# Qual: Adjust cache configuration for windows-ci
actions/cache@v4 is flawed on windows, try suggestions from issue
discussion to fix it.
https://github.com/actions/cache/issues/1361
.
* Make cache key depend on install path directory
* Qual: Adjust cache configuration for windows-ci
# Qual: Adjust cache configuration for windows-ci
actions/cache@v4 is flawed on windows, try suggestions from issue
discussion to fix it.
https://github.com/actions/cache/issues/1361
.
* Make cache key depend on install path directory
* QUAL: Add sqlfluff (SQL code and style check)
# QUAL: Add sqlfluff (SQL code and style check)
This adds a validity and style check on the .sql files.
The same tool can be used to fix style (which can be set up as a
pre-commit hook).
* Ignore some sqlfluff notices
* Ignore RF04 notice, warning about the use of sql keywords
* Adjust dialects for some directories
# Qual: Bump logToCheckStyle for better filename handling
Some notifications from the PHPCS step in pre-commit extracted too many
characters for the filename in a specific case.
The update to logToCheckStyle fixes that.
# Qual: Always run phpcs in pre-commit workflow
Developers are prioritizing the Travis Run because the pre-commit workflow
only runs the phpcs checks if the other pre-commit checks did not fail.
This modifies the action so that the phpcs checks are also run if the
previous step failed.
That should help limit travis load and delays.
# Qual/Fix Win-CI Check that the PHPUNIT step really completed
The program under test can abruptly abort the PHPUNIT test case by calling
exit which results in no error code - it then appears as if the test was
successful.
This adds a grep on the last lines looking for the summary information from
phpunit.
If absent, the step fails
# Qual: Exclude some tests from windows-ci until fixed
Until #28598 and #28264 find some way to the develop branch, this update will
skip the tests that currently fail on windows so that the results of other
tests have an effect on the CI status and github annotations
* Fix phpstan workflow file
# Fix phpstan workflow file
Line was split in two, joined
* qual: set memory limit to 'unlimited' (-1)
* Remove some stubs to check impact
* Fix: Verify execution time limit
* Remove stub files to reuse cache
* Qual: Rename distribution phpstan.neon to phpstan.neon.dist
# Qual: Rename distribution phpstan.neon to phpstan.neon.dist
This allows the developer to override some settings locally
Also update the .gitignore file with phpstan.neon and some other other
ignores (cache, local log files, temporary edit files).
* Enable debug on phpstan
* Re-enable stubs
* Fix: Change parallel settings to defaults, remove buffer
'buffer' is not a documented parameter, removing it.
Removing some configuration settings as the defaults are higher and
should adapt better to the target machine
# Qual: Clean up caches when PR is closed
This will clean up caches when a PR is closed.
This requires that `actions:write` is enabled.
Co-authored-by: Laurent Destailleur <eldy@destailleur.fr>
* Qual: ci: Run pre-commit/php-cs with cache (#28079)
This adds a hook to .pre-commit-config.yaml and updates the workflow
to run php-cs with cache when it is run for all files.
When running on changed files only, the cache is not useful.
The php-codesniffer ruleset.xml was cleaned up (duplicates removal/formatted)
* Fix: Make all 'relative paths' absolute (#28196)
# Fix: Make all 'relative paths' absolute
The phpcs ruleset xml file's relative exclude patterns are relative to
the filename(s) provided on the command line.
Hence with partial verifications, the path exclusion does not function
as we would like.
Removing the relative-path attribute from the patterns the exclusion
works.
At the same time, the patterns were optimized and a comment was added.
* Qual: Optimize workflow (#28386)
# Qual: Optimize workflow
The log annotation based on the pre-commit logs is now simplified.
* Fix: Workaround for false security issue in tests
# Fix: Workaround for false security issue in tests
This is a quick workaround for the false security issues in tests
by setting the ['SELF'] variable to something that is
considered harmless
* Update windows-ci workflow
# Update windows-ci workflow
- Setup_conf.sh - Added caching feature;
- phpunit.bat - Example updated based on experience with php web server
- Provide dolibarr.log and phpunit output as artefacts.
- Convert PHPUNIT log to Github Annotations
- Setup php web server for API tests
- Enable caching sql initialisation
* Qual: Re-enable open_basedir restriction in tests
# Qual: Re-enable open_basedir restriction in tests
Using open_basedir restriction is closer to the production reality
and seemed to generate some issues on a windows development machine.
---------
Co-authored-by: Laurent Destailleur <eldy@destailleur.fr>
* Dev: Add setup scripts for phpunit/database
# Dev: Add setup scripts for phpunit/database
Some scripts that helped to setup some unit testing.
* NEW Add Windows ci workflow
# NEW Add Windows ci workflow
This adds a workflow to execute ci on a windows platform
* Disable open_basedir for now, to test effect
* Qual: Exclude documents directory from codespell
# Qual: Exclude documents directory from codespell
Documents directory may be present during development (unit tests).
Do not check spelling there
* Qual: Add codespell exceptions
# Qual: Add codespell exceptions
Add new exceptions to codespell (var identified as misspelling).
# Qual: Improve caching for phpstan
The phpstan step does not seem to always choose the best cache.
This configuration update adds the target branch to the restore
feature.
Maybe it is more selective and better.
* Qual: Lower severity on php-cs messages that should not block (#28072)
# Qual: Lower severity on php-cs messages that should not block
Some warning messages result in php-cs having an exit code of 2 and others
only have an exit ocde of 1. When the exit code is 2, the warnings are fixable
by php-cbf.
In order to be able to filter on the warnings in the github ci flow, but still
show them in the travis flow, I lower the level of the warnings that should not
be blocking to 4.
At the same time, the php-cs configuration in pre-commit now requires that the
severity is minimum 5 to show the message (level 5 is the default for all warnings
and errors).
So this will limit the warning messages to the ones that should result in a failing
ci action. If any new cases are discovered, they should also be lowered in the
php-cs configuration file.
* Qual: ci: Run pre-commit/php-cs with cache (#28079)
This adds a hook to .pre-commit-config.yaml and updates the workflow
to run php-cs with cache when it is run for all files.
When running on changed files only, the cache is not useful.
The php-codesniffer ruleset.xml was cleaned up (duplicates removal/formatted)
This adds a hook to .pre-commit-config.yaml and updates the workflow
to run php-cs with cache when it is run for all files.
When running on changed files only, the cache is not useful.
The php-codesniffer ruleset.xml was cleaned up (duplicates removal/formatted)
# Qual: disable PHPCS action (now performed by pre-commit)
The php-cs analysis is now ensured by pre-commit which does analysis
on changed files in PR's and full analysis in the main branches.
It also provides better reports (PHPCS did not report all the issues
seen in travis).
This change to the workflow will run phpcs on all files in develop,
17.0, 18.0, ... (rule== branch ends in .0).
This ensures that phpcs is run in a github workflow for fast phpcs
issue detection in the develop and other main branches.
Ultimately, this means that the PHPCS action can be disabled
in branches where it is active. But I propose to verify that
the updated action does the job as expected (tested in a fork
it is ok).
Co-authored-by: Laurent Destailleur <eldy@destailleur.fr>
This change to the workflow will run phpcs on all files in develop,
17.0, 18.0, ... (rule== branch ends in .0).
This ensures that phpcs is run in a github workflow for fast phpcs
issue detection in the develop and other main branches.
Ultimately, this means that the PHPCS action can be disabled
in branches where it is active. But I propose to verify that
the updated action does the job as expected (tested in a fork
it is ok).
Backport the pre-commit tool to bring extra checks to the older versions
so that they are fixed earlier before merging in upper versions.
Other PRs will be created for backporting to 18.0 and 19.0.
The PRs should be accepted in reverse order:
- develop;
- 19.0:
- 18.0;
- 17.0.
That should avoid merge conflicts and ensure that options (and ignored
messages or steps) are adapted to the version branch.
* Qual: Backport pre-commit to 17.0
Backport the pre-commit tool to bring extra checks to the older versions
so that they are fixed earlier before merging in upper versions.
Other PRs will be created for backporting to 18.0 and 19.0.
The PRs should be accepted in reverse order:
- develop;
- 19.0:
- 18.0;
- 17.0.
That should avoid merge conflicts and ensure that options (and ignored
messages or steps) are adapted to the version branch.
* Qual: Backport pre-commit to 18.0
# Qual: Backport pre-commit to 18.0
Backport the pre-commit tool to bring extra checks to the older versions
so that they are fixed earlier before merging in upper versions.
Other PRs will be created for backporting to 18.0 and 19.0.
The PRs should be accepted in reverse order:
- develop;
- 19.0:
- 18.0;
- 17.0.
That should avoid merge conflicts and ensure that options (and ignored
messages or steps) are adapted to the version branch.
* Qual: Backport pre-commit to 19.0
# Qual: Backport pre-commit to 19.0
Backport the pre-commit tool to bring extra checks to the older versions
so that they are fixed earlier before merging in upper versions.
Other PRs will be created for backporting to 18.0 and 19.0.
The PRs should be accepted in reverse order:
- develop;
- 19.0:
- 18.0;
- 17.0.
That should avoid merge conflicts and ensure that options (and ignored
messages or steps) are adapted to the version branch.
# Qual: Optimize/explain workflows
phpstan checkout took about 1.5 minutes, while pre-commit only takes 10s.
So I removed the option to the checkout option which likely makes
the difference.
I removed the IMHO useless step of installing node - it may have been
useful during setting up the workflow, but not now.
I commented on the pre-commit workflow.
The PHPStan steps run a long time for the latest runs and I suspect that
this may be because of a bad cache, possibly includeing:
```
Result cache was not saved because of non-ignorable exception: Syntax error, unexpected } on line 378
``` .
This change should only save the cache if the run was successful or if no cache
was loaded. The possibly preserves a cache that still reduces the run time.
For instance, the following got an empty cache, saved at 2024-01-19T21:28:22.18Z.
- https://github.com/Dolibarr/dolibarr/actions/runs/7589384926
That was an aborted run:
- https://github.com/Dolibarr/dolibarr/actions/runs/7589372703/job/20673878193
So I added `cancelled()` to avoid overwriting valid cache and phpstan cache
will only be written if the outcome was success or if the cache could not
be loaded before.
I also updated the version for the cache action.
# Qual: Fix line endings of yaml files for yamllint.
yamllint (included in the pre-commit setup) has some messages about the line-endings of yml files.
This fixes that.
# Qual: Fix trailing whitespace in yaml file.
Yaml check tools complain about trailing whitespace, so I fix them.
pre-commit (#27392) has rules to fix them automatically (locally).
