[Date Prev][Date Next] [Chronological] [Thread] [Top]

Re: (ITS#6076) slapo-collect: slapd repeatable segfaults caused by deletion of a single olcCollectInfo config value (when olcCollectInfo contains multiple configuration values)



Hallvard B Furuseth schrieb:
> Yes, the on-th-fly sorting combined with the delete-single-value
> code seems broken.  I'm not sure if delete-single-value could ever
> be invoked 
Of course could it be "invoked" before: just use slapd 2.4.16 together 
with slapo-collect and your favorit ldapmodify client. Right after the 
"invocation" you'll get an error message complaining about a missing 
equality rule.
Thus there's no hint within collect's source that the single-value 
delete is faulty I inserted an equality rule and began to wonder...

> before you inserted an EQUALITY matching rule though.
Do you mean the single-value-delete functionality has been silently 
"disabled" by removing the EQUALITY matching rule leaving 
slapo-collect's by design defective single-value-delete code block 
knowingly unchanged/uncommented?
This seems very strange to me even more because slapo-collect is 
advertises as demonstration overlay (ok, a "quick hacked" one but 
nevertheless questionable).

> I too don't know cn=config all that well:-(
<speculation>
Instead of correcting handling this issue only in collect.c there 
perhabs exists a more general possibility to pimp the cn=config internal 
add mechanisms regarding the handling of ordered config values? In my 
opinion a very important feature concerning "overlapping"...

Á la:
cfg_value_add(.., .., SLAPD_CFG_VALSORT_ALPHA|SLAPD_CFG_VALSORT_DN|..)

only problem seems to be how to sort a configuration value like: 
'"ou=test,dc=foo,dc=bar" l,st' by DN's depth
</speculation>