mysqli: metacolumns() returning different info.

[NathanGibbs3]

Description

The metacolumns() function is returning different info in the master branch compared to the last release.

Environment

• ADOdb version: d4f9f45
• Driver or Module: mysqli
• Database type and version: MySQL 10.3.23 - 14.14
• PHP version: 7.0.25 - 8.1.12
• Platform: Linux

• I have tested that the problem is reproducible in the master branch

Steps to reproduce

Detailed, step-by-step instructions to reproduce the behavior, including:

• Code to reproduce & compare with 5.22.5

$lib = 0; // 0 use 5.22.5 release 1 use snapshot of master branch.
$driver = 'mysqli';

$svr = 'db';
$usr = 'test';
$pwd = 'password';

$DB = 'test';
$tbl = 'test';
$fld = 'test';

if ( $lib == 0 ){
    include('../../ADOdb-5.22.5/adodb.inc.php');
}else{
    include('../../ADOdb/adodb.inc.php');
}

$db = newAdoConnection($driver);
$db->connect($svr, $usr, $pwd, $DB);
$mctmp = $db->metacolumns($tbl);

print $db->Version()."\n";
print $db->isConnected()."\n";
print $db->ErrorMsg()."\n";
var_dump($mctmp[strtoupper($fld)]->max_length);

• SQL to create and populate database objects used by the code snippet

CREATE TABLE test ( test VARCHAR(255) );

Expected behavior

master branch code returns 255.

Additional context

Found this while working on #978
Reproduced locally and in CI.

[dregad]

Thanks for the bug report.

According to git bisect, the culprit is 5c9e24c. I'll look into it.

[dregad]

So the problem is caused by using $mode parameter instead of class property $adodbFetchMode (set by parent constructor). Problem is also present in other drivers, will add relevant tags.

[dregad]

Removing the 5.23.0 milestone, since the commit that introduced the regression was never part of an official release.

[NathanGibbs3]

Thanks for the bug report.

According to git bisect, the culprit is 5c9e24c. I'll look into it.

Oh, one of those. 😄
Fix an overarching issue in the app's architecture; then the wheels fall off everywhere else.
Usually because the app made assumptions based on the now corrected bug, or compensated for it in some way that is no longer applicable.
I've done it too, fix one bug, find 10 more. as a result. In the end the codebase gets better. 👍

[NathanGibbs3]

It looks like the parent class $fetchMode variable is not propagating down to the drivers.
Probably this at the beginning of the drivers will set things right.

if ($mode === false) {
	$mode = $this->fetchMode;
}

Unless the $mode flag for the drivers doesn't need to be passed anymore in which case, simply

$mode = $this->fetchMode;

without the if block should do it.

[NathanGibbs3]

Verified locally, both approaches to this issue correct it.
If you want, I'll go through and add the fix to the drivers changed in the git bisect above & put it together into a PR for you.
Would you want the single line or if block fix?

If you want the single line fix, is the $mode parameter for the drivers even necessary anymore?

[dregad]

Thanks for the offer, I have a patch ready that I'll push shortly. Got sidetracked as I started looking at your other issue 😉