Open staabm opened 8 months ago
I just found #4314 - which described how it should be done.
I was about to comment this:
$ ./tools/phpunit --filter foo --fail-on-empty-test-suite
PHPUnit 11.0.3 by Sebastian Bergmann and contributors.
Runtime: PHP 8.3.3
Configuration: /usr/local/src/raytracer/phpunit.xml
No tests executed!
$ echo $?
1
I just opened #5707 as --fail-on-empty-test-suite
is missing from --help
output.
We're just interested in how the value passed to --filter
should be sanitized when it comes from an unknown/dynamic source. Is doubling each \
the only thing that needs to be done, or are there other rules to follow?
I honestly do not know and need to research this (ie. look at the code). I do not have time to do that right now, sorry. I might get to it later today, but it's the weekend, you know. TL;DR: I will do my best to get you an answer ASAP.
There's no rush, we're just trying to improve the CI pipeline :) Thank you very much.
Doubling \
should be enough. Here you can see what PHPUnit does with the value passed to --filter
.
hey sebastian,
thx for doing the research.
I think we need to do more escaping, see the very first example in the PR description above.
when using data-providers with a string-key, its possible to use characters like e.g. $
, (
, )
which need escaping in the --filter
option.
Doubling \
is only enough if your filter just contains a class-name and or test-method, but not identifier for the data-provider.
Do you need something else from me here or can this be closed?
I think what @staabm tried to say is that when you use a custom name for a specific data provider iteration like:
yield 'unknown filter' => [
'mixed',
'filter_var($mixed, $mixed)',
];
You could use a custom name that looks like a regular expression. And the question is - if we wanted to execute only this test, how should we write the argument for --filter
so that it's not interpreted as a regular expression, but as a request to execute the test for this one specific data provider iteration.
@staabm's theory is that simply doubling \\
is not enough to properly escape the --filter
value.
We could of course avoid this problem by using simple names for data providers (and I think we do), but if we were for example to publish a general GitHub Action that would promise you to partition your test suite and run it in parallel over multiple jobs, we could not rely on simple names anymore.
@staabm In my opinion source of the problem is function preg_quote
which you consider for escaping data extracted from --list-tests-xml
.
In preg_quote
documentation https://www.php.net/manual/en/function.preg-quote.php we can read, that #
is "special regular expression character" and will be escaped by this function.
It surprises me, because in PCRE documentation https://www.php.net/manual/en/regexp.reference.meta.php #
is not present on meta-character list. Exactly the same we can find in https://www.pcre.org/original/doc/html/pcrepattern.html#SEC4
I think you should consider using str_replace
instead of preg_quote
.
Summary
we try to feed
phpunit --filter
with test-case names which we determine via--list-tests-xml
. we are wondering, how to properly escape our input strings we want to pass to phpunit via--filter
option.Current behavior
the docs mention the filter is interpreted as a regex, therefore I went with
preg_quote()
, but this didn't work when we use a "full qualified test-case name" which contains a#
.How to reproduce
a test-class like
requires a escaped filter
the escaping required looks like
preg_quote
and it works for the case abovethe following example doesn't work with
preg_quote
:Expected behavior
1)
I wonder how to properly escape the value for the
--filter
option. this would be especially important as when using the wrong filter no tests would be executed.2)
can we make phpunit return with a non-zero exit code, when a filter is used which does not execute any tests?edit: I just found https://github.com/sebastianbergmann/phpunit/pull/4314 - which described how it should be done.