| Age | Commit message (Collapse) | Author | Files | Lines |
|---|
|
Always clone, but left in mysqli check for PHP7
|
|
|
|
If ADODB_FETCH_MODE is ADODB_FETCH_BOTH, and 2 columns were returned,
the first column was returned as both the key and the value.
Fixes #600
|
|
Fixes #445
Signed-off-by: Damien Regad <dregad@mantisbt.org>
Rebased, conflicts resolved, fixed whitespace (indent with spaces).
|
|
|
|
This is @laurent-n's original submission from PR #489 with some
whitespace adjustments, rebased to latest master.
Signed-off-by: Damien Regad <dregad@mantisbt.org>
|
|
Microsoft wraps column names in square brackets in T-SQL statements,
which is not really obligatory but becomes necessary when reserved
keywords are used as columns names (e.g. 'BEGIN' or 'CURRENT').
Fixes #246
Signed-off-by: Damien Regad <dregad@mantisbt.org>
Original commits from PR squashed and message reworded.
|
|
# Conflicts:
# adodb-lib.inc.php
|
|
The new function is the same as getMenu3(), except that it has an extra
parameter $compareFirstCol to align it with getMenu(). The new param
defaults to true, i.e. defaults are compared against the value, not the
description (opposite of getMenu3()).
getMenu3() is marked as deprecated and just calls the new function.
|
|
Coding guidelines / whitespace.
|
|
It has been noted as such in the documentation since 01.01.2019, but
the code had not been updated accordingly.
|
|
Coding guidelines / whitespace.
|
|
|
|
|
|
|
|
|
|
|
|
|
|
srand() should not be called manually, unless one has a better way to
initialise the already-initialised random seed.
The way the current srand() call is made is buggy, and will lead to
duplicate random seeds if called often enough. Because the current call
will lose the timestamp part, and will only use the first 6 chars from
the current microsecond, but the system might not have a high-enough
timer resolution, so the actual pool of random seeds gets even smaller.
Fixes #532
Signed-off-by: Damien Regad <dregad@mantisbt.org>
|
|
When initializing a new connection with postgres or pgsql driver, ADOdb
currently defaults to using the postgres8 driver.
Postgres 8.4 lifecycle ended in 2014, and currently only versions >= 9.4
are supported, so it makes sense to change our default to connect using
the postgres9 driver instead of postgres8, even though they currently
only differ by driver name.
Fixes #474
|
|
|
|
# Conflicts:
# adodb-time.inc.php
# docs/changelog.md
# drivers/adodb-mssql.inc.php
# drivers/adodb-mssqlnative.inc.php
# drivers/adodb-oci8po.inc.php
|
|
|
|
|
|
|
|
|
|
|
|
Commit 53d13c0f64ad97ff87fa3b74fa4a4e896e691c9b (fixing #58 and #63)
introduced a regression in 5.20.0 that broke connections LDAP when using
URIs like ldap:// or ldaps://.
Create a new protected method ADOConnection::parseHostNameAndPort() to
help parse off the port from a hostname while retaining the ldap:// or
ldaps:// scheme, and use it to refactor Connect() and PConnect().
Fixes #340
Signed-off-by: Damien Regad <dregad@mantisbt.org>
The PR's original commits were squashed and the message reworded. The
fix (initially targeted at master) was applied to hotfix/5.20 branch.
|
|
|
|
Conflicts:
adodb.inc.php
docs/changelog.md
|
|
The PHP native `$php_errormsg` variable has been deprecated as of PHP
7.2. The recommended replacement is `error_get_last()`.
A number of the driver classes used `$php_errormsg` to track if a DB
action caused an error.
As the `error_get_last()` function returns an array and comparing the
message therefore is slightly more involved, two new methods have been
introduced into the abstract parent class `ADOConnection` to handle the
message comparison.
The `resetLastError()` method retrieves the initial (old) error message
and resets it to enable checking whether a new error has occurred, even
when the new error is the same as the previous one.
The new `getChangedErrorMsg()` method compares the old error message
with the current one to determine whether a new message has occurred,
and if so, returns the message.
Refs:
* http://php.net/manual/en/migration72.deprecated.php#migration72.deprecated.track_errors-and-php_errormsg
* http://php.net/manual/en/reserved.variables.phperrormsg.php
* http://php.net/manual/en/function.error-get-last.php
Additional notes:
* In the `ADODB_ads::UpdateBlob()` method, the original (old) message
was not being retrieved which could result in an unrelated error being
recorded. This has been fixed.
* In the `ADODB_db2::_connect()` and `ADODB_db2::_pconnect()` methods,
the `$php_errormsg` was being reset, but subsequently not checked for
changes, so as the variable in effect was not being used, I've removed
the related code.
* In the `ADODB_sybase::ErrorMsg()` method, the `$php_errormsg` was
being checked, but no functions which could potential cause errors are
being called. As `$php_errormsg` is only available in the scope an
error occurred in, this meant that `$php_errormsg` would always be
empty, so I've removed the related code.
Fixes #409
|
|
|
|
Cherry-pick + squash of commits
- 5c5ee9e953dd0caf978630b1c3147447d6cb58a7
- c8524bb3fcd291fe6233643fb43e9ceeb8aef925
|
|
Backported from d9cc6c66ec08e5a517f8fa4503345ead992e2a62. I meant to do
this in 5.20.10 but I forgot...
Fixes #373
|
|
|
|
|
|
|
|
To avoid unwanted information disclosure if an error occurs when
loading a database driver, we should not output the full path in the
error message.
Fixes #389
|
|
https://pecl.php.net/package/memcached
Fixes #322
|
|
|
|
We had a problem in our app using the pdo_sqlsrv driver with PHP 7 that
manifested itself by a ODBC32.dll crash at the end of a PHP request
(see issue: https://github.com/Microsoft/msphpsql/issues/434).
It seems the dll tried to free a statement after the connection had been
disconnected.
We looked at our code and the ADOdb code to see if a reference to a
statement could imply that it would be garbage-collected after the PDO
disconnection.
We found that `ADODB_pdo#_query` returns a statement to
`ADOConnection#_Execute` which stores it in its `_queryID` property and
never empties it explicitly.
Setting it to `false` in `ADOConnection#Close` seems to fix the problem
in our app.
Fixes #379
|
|
Currently when defining $ADODB_OUTP you can set a custom output
function. However i want to use an object method as an output handler.
So now $ADODB_OUTP is fed to call_user_func so we can set:
$ADODB_OUTP=array($object,'output_method');
Fixes #312
|
|
|
|
|
|
|
|
Most of the includes in adodb.inc.php are done with include_once but
in a couple locations its still done with include.
This is the case for adodb-lib.inc.php which is conditionally included
based on a global var. I ran into problems with running adodb in debug
mode while running phpunit tests. Phpunit will reset the global vars
and thus break the include mechanism. This is a problem of phpunit
and can be worked around. However it would make adodb more robust
to just use include_once everywhere. It will remove the dependency on
the $ADODB_INCLUDED_LIB global var.
I also replaced one include statement in adodb-time.inc.php, in various
drivers and other files too, so include_once is used everywhere.
Signed-off-by: Damien Regad <dregad@mantisbt.org>
|
|
|
|
see #350
|
|
Changed tabs in annotations to spaces
Fixes #345
|
|
Fixes #344
|
