We are running a *very* modified version of htdig-3.0.8b1, the code has been heavily modified to do many extra things but these modifications mean that we cannot easily upgrade to htdig-3.1. We have hit what we are told are gdbm size limitations of 2GB. Running htmerge fails in the last part of the merge (where it is counting up the documents it has seen) with "Virtual memory exceeded in `new' ". We are not out of memory as we have 5GB of swap and vmstat shows that we are only using about 2.1GB of it. htmerge... htmerge: Sorting... htmerge: Removing doc #0 htmerge: Merging... htmerge: 312800:
�stensj�veie htmerge: Total word count: 312867 Virtual memory exceeded in `new' We are running under Solaris 2.6 on a Sun Enterprise 450/2250. We are indexing about 550,000 documents, our file sizes (at the time the merge failed) are: 957969723 Oct 27 19:34 db.docdb 67426370 Nov 4 21:48 db.docs.index 953222215 Nov 4 17:43 db.wordlist 532695286 Nov 4 17:43 db.words.gdbm To fix our current problem (and to get better performance) we would like to change from gdbm to the Berkeley db but we cannot do it by applying the patches as they were made against a standard system and ours is anything but that. Could someone give us some pointers as to which parts of the code we would need to change in order to use the Berkeley db. Also any help with the current problem as a short term fix would be much appreciated. Thanks in advance, Paul Lucas
