Improve DBA test suite

[Girgias]

Main drive for this is to make the DBA tests parallel and remove redundant tests that are now covered in a large suite for all drivers.

I don't have the db1, db2, db3, dbm and ndbm drivers so those are best guesses for the expectation of the tests.

Other objectives of this improved suite is to make the DB4 test which tests opening of a DB run on all the other drivers, as this has found an issue with LMDB.

[cmb69]

I don't have the db1, db2, db3,

Does anybody else? I suggested to drop support for these three some years ago, but nothing came out of it. Unfortunately, some links in this thread no longer present the relevant information, but if it is true that Berkeley DB 3.3.11 was the final release of the 3.XX versions, and has been released on 2001-07-12, it appears to be silly to spend any effort on that; except for dropping support.

[Girgias]

I don't have the db1, db2, db3,

Does anybody else? I suggested to drop support for these three some years ago, but nothing came out of it. Unfortunately, some links in this thread no longer present the relevant information, but if it is true that Berkeley DB 3.3.11 was the final release of the 3.XX versions, and has been released on 2001-07-12, it appears to be silly to spend any effort on that; except for dropping support.

Right, that's what I'm also thinking, I don't think we can just drop support in 8.2 especially as we don't have time for an RFC :/

[Girgias]

@cmb69 I don't understand why I'm getting a segfault on Windows:

========DIFF========
--
     Mode parameter is "cd":
     This is a test insert
     Mode parameter is "c-":
039+ 
040+ Termsig=-1073741819
039- This is a test insert

For FAIL DBA flatfile opening matrix of combination [C:\projects\php-src\ext\dba\tests\dba_flatfile_creation_matrix.phpt] (and 2 other tests which are similar)`` I though maybe 5995d3c would do something to help it but seems not.

Could you have a look?

[cmb69]

It seems you've found a (long standing) bug. Simple reproducer: dba_open("non-existing-file", "c-", "flatfile"). The problem is that we're trying to free a NULL pointer:

php-src/ext/dba/dba.c

Line 830 in 2fd5e82

php_stream_free(info->lock.fp, persistent ? PHP_STREAM_FREE_CLOSE_PERSISTENT : PHP_STREAM_FREE_CLOSE);

The simplest solution would be to only free if info->lock.fp is set. However, there is no need to close the streams and to restart if we don't want to lock at all. This could be caught in the else if clause a few lines above. And frankly, I'm not sure whether this restarting is necessary nowadays.

Anyhow, I'll come up with a fix, and suggest you ignore that problem for now. It might also be good to revert 5995d3c.

PS: after closer investigation, I found that I misinterpreted the restart.