The problem still persists, but I have found a workaround now: when I
tried generating the index on a system with an ancient Debian 4.0, there
is no segfault problem, and fortunately the Debian 7.1 swish-e reads
that index fine, when copied.
When I copy the Debian 7.1 generated index to the Debian 4.0 system, the
segfault persists, so I guess it is a writing issue on Debian 7.1.
If I keep around the Debian 4.0 machine, index generation should work,
but it is not optimal and some kind of solution for 64bit Debian 7 would
probably be good: I ran about 45000 queries with extracted words against
the 64bit machine with the problem index, resulting in a 5% segfault
rate, which I find quite substantial.
Anyone have an idea what could be the problem? I have made my collected
data available under
(about 426KB, 2.6MB uncompressed). It contains a README file describing
the contained files.
Users mailing list
Received on Wed Nov 06 2013 - 10:27:00 GMT