PR #3852 was a bit strange. Php-cs-fixer complained about a problem that I could not reproduce on my local system, and it didn't really produce sufficient output to correct whatever the problem was. I was able to guess it eventually, but I need to see if there's a way to get it to be more forthcoming.
PHPDoc types can sometimes be entirely expressed as PHP native types. It
is better because it avoids code duplication and enables type runtime
check.
This will help us slowly migrate away from PHPDoc typing to PHP native typing.
This is according to our formal, published, policy to only support
EOL PHP after 6 months.
See https://phpspreadsheet.readthedocs.io/en/latest/#php-version-support
Also share the exact same dev deps across all PHP version for GitHub
Actions so runs are faster, much most importantly they are stable and
predictable. And we decide manually when we want to migrate to PHPUnit
10.
Fixes#3634Closes#3710
* PhpUnit 10 Compatibility Part 3 (Last)
Final changes for PhpUnit 10, including enabling it for testing. This finishes the work of PR #3523 and PR #3526.
The major remaining problem with PhpUnit 10 is that earlier releases converted notices and warnings to exceptions, and 10 does not. Not having the information provided by the messages seems risky to me. Fortunately, it appears that you can add an error handler to the test bootstrap for 10 and make it act like earlier versions; I have done so. In order to demonstrate the effectiveness of this handler, a new otherwise unused class Helper/Handler and tests for that class are added.
As part of the testing of this change, it became apparent that the fopen in OLE::getStream attempts to create a dynamic property $context in Shared/OLE/ChainedBlockStream, and that action is deprecated in Php8.2. Adding the property to the class eliminates that problem. No executable code is added, and this is the only change to source code.
There also seems to have been a change in assertXmlStringEqualsXmlString in PhpUnit 10. The only test which uses that method is Chart/Issue589Test, and both the places which use that method could just as easily and effectively use assertSame. They are changed to do so.
* Remove Phpunit Verbose Option
Not supported in PhpUnit 10. I'm not at all sure that this is the correct solution for this problem.
* Try Changing Phpunit Command for Different Php Releases
Not sure how to test this locally. We'll see if it works on github.
* Another main.yml attempt
We shall see.
* Eliminate One Test
Not sure why testDeprecated is not working in Github; it works locally. Disable it for now and continue to research.
* Show Incomplete and Other Messages in PhpUnit 10
6 new command line args to replace the 1 they got rid of.
* Restore Disabled Test
Deprecated messages are suppressed by default setting for error_reporting. Switch that to E_ALL in bootstrap and restore original test.
* Add Deprecation Tests for PhpUnit 9-
Default configuration option caused deprecation messages to be suppressed. Change the option.
* WIP Run phpcs, php-cs-fixer, phpstan, coverage, versions as Php8.1
They all run under Php7.4 in Github. 7.4 is EOL. We still have to run unit tests in 7.4, but I think it's time to move the tools. Note that we cannot currently run Phpstan in 8.2 because of https://github.com/phpstan/phpstan/issues/8629.
* Update main.yml
Try running coverage as 8.0 rather than 8.1.
* Update main.yml
Revert Coverage to Php 7.4.
* Composer Cache Directory
Command set-output is deprecated. Upgrading to using Environment files as suggested by Github messages.
* Coverage and Php8
Try to follow advice in https://github.com/scrutinizer-ci/ocular/issues/54
* Keep Trying
See https://github.com/phpowermove/docblock/pull/12/files
* Attempt to provide allow failure for PHP8.1 unit tests
PHP8.1 Tests show as passed despite the errors, and it requires checking the actual output from the run to see what the rea result is; but I can live with that until github provides functionality for a proper allow_failure option
The plan is to keep Travis for a short while, until we are confident that
GitHub Actions work well enough for us. And after that we can remove Travis
entirely.
There is a bunch of duplicated things but it allows us to maximize
parallelismt to have results as soon as possible.
API documentation generation is still missing.
chris
oleibman
Adrien Crivelli
Serhii Petrov
Alex
MarkBaker
MarkBaker
Sergiy Petrov