Fix #69592: pdo_dblib now skips datasets without columns while fetching data

[MiRacLe-RPZ]

Some sql-statements can return empty resultsets, now pdo_dblib skips empty datasets like ext/mssql.

[adambaratz]

I like this change, but there's probably code in the wild that relies on this (admittedly broken) behavior. It would make for smoother PHP upgrades if this was an opt-in behavior, via a driver attribute or whatever.

[MiRacLe-RPZ]

@adambaratz, several versions ago "code in the wild" cannot be relies on "that", because call of nextResult causes segfault (see PR #1386). This behaviour copied from ext/mssql(dblib).

[adambaratz]

I haven't dug into the code enough to understand why, but the segfaults weren't a consistent issue. My company has a fair bit of code that expects empty rowsets.

Would it be hard to add a boolean attribute to control this? It could be deprecated after a few minor revisions, but it would make it possible to transition out of this behavior instead of causing code to break when you upgrade PHP.

[MiRacLe-RPZ]

@adambaratz can you show testcase with code that expects empty rowset(s) ?

P.S. - a "empty rowset" i mean dataset without columns, not rows...

[adambaratz]

The gist would be to change your example like this:

$stmt = $db->query($sql);
$stmt->fetchAll(); // swallow empty rowset...
$stmt->nextRowset(); // ...and move to the next
var_dump($stmt->fetchAll(PDO::FETCH_ASSOC)); // expected: array(result => 0), actual: array()
var_dump($stmt->nextRowset()); // expected: bool(false), actual: bool(true)
var_dump($stmt->fetchAll(PDO::FETCH_ASSOC)); // expected: array(), actual: array(result => 0)

[MiRacLe-RPZ]

@adambaratz ok, i will try to add this functional optional by default, last question (programmer's main problem) - name for this pdo-attribute ?

[adambaratz]

It seems like the convention for driver-specific attrs is PDO::[DRIVER]ATTR[name]. I think PDO::DBLIB_ATTR_SKIP_EMPTY_ROWSETS would be fine.

[weltling]

Is there a 7.0 compatible patch for this issue?

Thanks.

[krakjoe]

Since this targets a security fix only branch, and since the author seems to have abandoned working on it (or responding to questions/comments), I'm closing this PR.

Please take this action as (more) encouragement to open a PR against a supported branch.