On Mon, 18 Mar 2002 16:17:08 GMT, I wrote:
> I realise this isn't much to go on, but would apprecaite pointers for
> other things that I need to check or try out.
Following up to my own message, to tie up the solution within the
archives. Unfortunately I had to wing this one on my own, and after a
week's worth of pain, I got there in the end.
I believe ultimately the problem was down to NFS, but the errors that I
was getting changed over time from the ones I originally posted to errors
stating that block x could not be written, and resource temporarily
unavailable - but with no other indicators. I must have tried about 30
different compilations on both 7.1.3 and 7.2.
It was unfortunate (for this) that my home directory is mounted onto many
machines and I didn't read anything about not running the database or
backend on an NFS partition.
So, the check list for Solaris (5.8) seems to be:-
1) Read doc/FAQ_Solaris, and ensure that tests are performed over a TCP
socket, and not a Unix domain socket.
2) Ensure that the shared memory and semaphore settings are high enough in
/etc/system
3) Build and run the regression tests on local disc storage, and not NFS.
Optimisation flags are optional, but I've left them in now for hopefully
faster response.
- Dave.
