If the content is invalid for whatever reason e.g. when instantiating
a DBText field to get a summary of text through
`DBField::create_field('Text', $content)->Summary(10)`, preg_split returns
false and the rest of the code expects an array.
This tweak ensures an array is always returned even when preg_split fails.
* DOCS Clearer sysadmin guidance for "packaging"
We have all kinds of fun fallbacks that attempt to create supporting files in production environments.
The latest point of contention is dev/build automatically creating files in .graphql/ and public/_graphql/
if those don't exist. That should be regarded as a last resort option to allow introduction of GraphQL v4 in the CMS 4.x release line.
At least since CMS 4.1, we need some form of "packaging" for generated files (public/_resources),
or committing these into the codebase, so let's call that out for anyone running CMS infra.
* Add trailing slash
Co-authored-by: Aaron Carlino <unclecheese@leftandmain.com>
- The Silverstripe 4 folder structure has been changed from **app/code/** to **app/src/**
- Renamed Silverstripe in text. I assume `api:` and `Namespace`, should remain SilverStripe
- Added some missing semicolons
This new opt-in setting will let grid field detail forms redirect to the
“Correct” URL of a GridField if it’s not found in the current list.
This works by:
* Looking for the item in the database
* If it exists, check for a CMSEditLink() method that returns a value
* If so, redirect to that
This is useful if you have a number of grid fields that each show a
partial list of records, and it’s possible for the user to make changes
such the item no longer appears in the list, but does appear in another
list.
It’s an opt-in feature as I think all changes like this should be
opt-in, based on previous experiences improving GridField and in turn
breaking SecurityAdmin and slowing versioned-data-browsing down. ;-)
The functioning of dot-syntax in form fields mean that .s are more
likely to appear in names. This breaks javascript behaviour in HTML IDs
and I believe is an invalid character for them.
Wrap doesn't actually wrap in the tidy extension.
This causes tests to be flakey, for example some of `FormTest` fails
if you happen to have `ext-tidy` installed (which is the default on most systems).
This happened to me on PHP 7.4.19 with tidy 5.6.0 (OSX Homebrew).
Note that the tests aren't failing in Travis right now.
You'd expect `wrap => 0` to be honoured. It's documented as an integer
in the tidy docs: https://api.html-tidy.org/tidy/quickref_5.6.0.html#wrap.
Even tracked this down to the PHP source which appears to be doing the right thing:
https://github.com/php/php-src/blob/master/ext/tidy/tidy.c#L300
There's a bug from 2018 against PHP 7.2.8 which was closed as "not a bug" without comment:
https://bugs.php.net/bug.php?id=76683
You can see the behaviour in action in the following test.
```
<?php
$html = '<p>a really long string which should not be wrapped</p>';
echo "## With tidy extension" . PHP_EOL;
$tidy = new tidy();
$tidy->parseString(
$html,
[
'output-xhtml' => true,
'numeric-entities' => true,
'wrap' => 0,
],
'utf8'
);
$tidy->cleanRepair();
echo $tidy . PHP_EOL;
echo "## With tidy cli" . PHP_EOL;
$cmd = sprintf("echo " . escapeshellarg($html) . " | tidy --force-output 1 -n -q -utf8 -asxhtml -w 0 2> /dev/null");
echo shell_exec($cmd);
```
Long story short, setting it to 99999 fixes the issue.
