RE: Slapd fails to die (ITS#2502)

[Villy Kruse <vek@pharmapartners.nl>]

On Fri, 6 Jun 2003, Howard Chu wrote:

>
> Seems like an fcntl lock on byte 0 of the id2entry database would be good
> enough.
>


Or whatever file locking primitive is supported by the system in question.

   System V  fcntl
   BSD       flock
   Linux     flock or fcntl
   NT        ?
   ?         ?

It seems that BerkeleyDB itself doesn't use any of these locking primitives
on that particular file, so you should not get any locking conflicts with
DB itself.  But perhaps someone should doublecheck that on systems where the
DB mutexes are implemented using file locking primitives.


Villy