Windows related changes, ahead/behind only current branch

[alfsb]

This PR makes languages.php more functional on Windows, and also changes repository status to not show ahead/behind of unrelated branches.

As a heads up, also remove a ugly hack for dirs named PDO vs pdo. If you are on Windows, and start receiving errors of undefined entities that contains PDO (upper case), you may need to filesystem remove all files from en/reference/pdo*, and then git restore then, as git on Windows may not honor case changing moves.

[cmb69]

I don't quite understand why you first replace backward slashes with forward slashes, and later call realpath() which will reintroduce the backslashes.

[alfsb]

Because realpath( "C:\phpdoc\en" ) will fail on Windows, I know that is because escaping and not directly related with realpath, but it really frustrating to have some literal "file part" sometimes behaving differently.

And to avoid code end looking like this:

const RNG_SCHEMA_DIR = __DIR__ . DIRECTORY_SEPARATOR . 'docbook' . DIRECTORY_SEPARATOR . 'docbook-v5.2-os' . DIRECTORY_SEPARATOR . 'rng' . DIRECTORY_SEPARATOR;

[alfsb]

I cannot test ir now (soon starting some new year holidays), but I vaguely remember that libxml will accept foward slashes on Windows, backward slashes, but not mixed slashes, e.g., C:\phpdoc/en.

[cmb69]

Well, yeah, you can usually just use forward slashes on Windows. But I still don't understand why you first replace backward slashes with forward slashes, and later call realpath() which will reintroduce the backslashes. I mean that has nothing to do with literal file paths, nor with mixing forward and backward slashes.

[alfsb]

I will have to test again, without this PR, to see exactly what was broken. I think there is something git related, as this function was copied from languages.php, that does a lot of git invocations.

[alfsb]

Remembered now. realpath() does not reintroduces foward slashes in all cases. It's not even guaranteed to return string. This function is called for expected non existent files, where realpath() returns bool, and I was under impression that this may fails on Windows, and at the time (as now) it's difficult for me to access Windows boxes to make tests. Yet. I remembered that forwards slashes always worked.

Also, these is some evidence this was the norm on doc-base scripts. xml-check.php does it, and I remember that previous file-entities.php there is a lot oft this too.

[cmb69]

doc-base/scripts/xml-check.php

Line 32 in 643f7dd

$realpath = str_replace('\\', '/', realpath($_SERVER["argv"][1]));

That code calls realpath() and afterwards replaces the backslashes with forward slashes. That can make sense. Even if the file does not exists, you get false and the str_replace() will return an empty string. Even that might make sense.

I still fail to understand, why you first replace forward slashes with backward slashes, and then call realpath() which will always use backward slashes on Windows (unless the file does not exist, in which case the given filepath will be returned with forward slashes). Maybe you want something like this:

{
    if ( $touch && ! file_exists( $path ) )
        touch( $path );

    $res = realpath( $path );
    if ($res !== false)
        $path = $res;

    return str_replace( "\\" , '/' , $path );
}

[alfsb]

I still fail to understand, why you first replace forward slashes with backward slashes

To avoid mixed slashes (C:\phpdoc/en) when realpath() fails. All backwards, all fowards, never mixed.

and then call realpath() which will always use backward slashes on Windows

The "foward slashes everywhere" refers to using foward slashes everywhere in source code. Having all slashes being foward on Windows when used (replaced), is better for compatibility.

That said, I will remove this line.