Finally, I could reproduce the problem (-m problem).
I am fixing it.
I will release beta4 in a few hours with the fix.
Another problem fixed in beta4:
This is related with performance. It is not a functional
When working with about 70000 files I noticed that
searching was very slow on solaris.
The test case shows that searching a word that was in just 4
files took about 12 seconds in my solaris installation. It had
no sense. This was much more slower than in AIX or Linux...
So I thought it was a solaris problem...
Tracking the code I discover that the function getindexfilenum
was very slow in solaris. I think this is because of the well known
bad I/O of SUN's boxes (the test was made on a 4 CPU SUN
4500 with 3GB of RAM!!!). Small AIX or LINUX boxes outperformed
SUN's box in this test case.
The workaround is very easy: getindexfilenum is a function for
counting the file entries of the index file but this number is also
in the file index header!!! So, reading all the file entries is not
This is also a problem of swish-e-1.X
Received on Thu Jul 27 05:28:27 2000