Re: Detecting an error from ldap_get_values()

["Kurt D. Zeilenga" <Kurt@OpenLDAP.org>]

At 11:20 AM 3/21/00 -0800, Kurt D. Zeilenga wrote:
>At 08:07 PM 3/21/00 +0100, Julio Sánchez Fernández wrote:
>>"Kurt D. Zeilenga" wrote:
>>> I'm thinking that the LDAP C API I-D should require the
>>> API to set LDAP_ERROR_NUMBER upon ldap_get_values() error.
>>> This issue should be raised to IETF-LDAPext.
>>> 
>>> In the mean time, I would not object to adding code to
>>> OpenLDAP-devel to set LDAP_ERROR_NUMBER.
>>
>>It is not setting, it is clearing a prior error code.
>
>I see.  Then the caller (sendmail) should clear the error code
>before making the call.  The API is only allowed to set LDAP_ERROR_NUMBER
>with the server provided resultCode or an code reserved for API use.
>It's not allowed set the session LDAP_ERROR_NUMBER to LDAP_SUCCESS.
>
>Sounds more and more like an IETF LDAPext WG issue.

s/LDAP_ERROR_NUMBER/LDAP_OPT_ERROR_NUMBER/

In addition:

We should verify LDAP_OPT_ERROR_NUMBER is set with the resultCode
from a response message OR, upon API error, to a value reserved
for API use.  There may be cases in the code where the API
inadvertly clears LDAP_OPT_ERROR_NUMBER.

In particular:
	ec = ldap_xxxx();

	ldap->ld_errno = ec;
	return ec;

constructs are broken.  ldap_xxxx should set ld_errno if an error
occurred, so:
	ec = ldap_xxxx();
	return ec;

should be sufficient.  If ldap_xxxx doesn't act upon a session
handle then the code should be:
	ec = ldap_xxxx();
	if( ec != LDAP_SUCCESS )
		ldap->ld_errno = ec;
	return ec;

Kurt