I'm not sure about #1, that sounds like a problem with the indexing of content which I wouldn't expect to be influenced by the db layer. Is the directory actually missing? What happens if you try a search, does the directory get created then? It should have been created during setup as content was created.

For #2 that was certainly my bad, apparently when I added that stored procedure to the script for postgre I thought I was editing my svn working copy but must have been editing a copy on my linux machine (I run pg from linux on a different machine) so the updated script never made it to svn until today and wasn't in the 2.1.8 release files for pgsql. I have just now corrected the downloads and versioned the pgsql release files as 2-1-8-1.

Thanks,

Joe

Well the bug seems to me to be that the index is not being created when the content is created during setup. When you do a search for the first time if the index is not there it indexes all the site content, so doing the search is what fixed it.

Probably the solution is to catch and handle the exception when it tries to delete from a missing index. Its really not a problem for it to not be there right away since it will be created upon the first search.

I would think this bug would happen using any data layer, so it seems odd if its only happening with pgsql.
If you have time, it would be good to know if the index is created immediately using MS SQL or if the same sequence of events can be produced. Its not really all that important though as the solution mentioned above will take care of it.

Thanks,

Joe