RE: [PORTS] RE: PostgreSQL on Win32

> >> I've tried to get PostgreSQL running on Windows NT using
> >> CygWin as described in the PostgreSQL documentation with
> >> quite a few problems, some relating to
> >> Cygwin, some to Postgres itself.
>
> > There was one big problem with freezeing the backend caused
> > by the cygipc library which could be fixed now.
>
> I'm running cygipc v1.03. Is this the buggy one? It looks
> like it hasn't
> been modified in a while.

I am using 1.03 too. There is a bug in initializations of global variables
in new processes. I have included the patch (it is not my work). More
details are in pgsql-hackers in the thread with subject "backend freezing
...".

>
> I've actually successfully compiled and run postgreSQL on
> windows NT... RTML
> worked on the third try <grin>.
>
> So now I've got it up and running under Cygwin b20.1 on a
> P166 w/ 64 megs
> RAM and I'm using the Insight ODBC drivers (v6.40.0006) to
> connect Access to
> it on the same machine.
>
> In preliminary testing, with one table (35,000 records) I'm
> finding it quite
> slow, possibly too slow to actually use in production. However, I'm
> wondering if this could be due to any of the following:
>
> [1] Running the db and the client on the same machine. (I
> only have one NT
> machine available at the moment so that's my development platform.)
>
> [2] Cygwin is slow (although psql client queries seem to be quite
> acceptable).

P166 is not very fast machine for Windows NT ;-) . My experience says that
compilation PostgreSQL in Linux is more than 2 times faster than in Cygwin
on the same machine.

>
> [3] The Insight ODBC driver.

I think that ODBC is generally slower than using native API of the database
server. I am developing a general API in C for accessing database servers
(something like DBI for Perl).

>
> [4] I've no indices on my test table.

Good index can speedup the queries.

>
> My testing consists of making a copy of an existing table in
> Access and
> doing "stuff" to the original table and the copied
> (PostgreSQL/ODBC) table.
> "Stuff" is simple things like inserting a record, deleting a
> record (which
> fails because another user allegedly modifies it), modifying
> data, sorting
> the table, etc. I think that point [4] would only come into
> play for the
> sorting part.
>
>
> > In my opinion the cygwin port is stable enough (in the core
> > functionality tested by the regression tests) now but it is
> > a relatively new port and the last known problem is being
> > fixed right now.
>
> Quite a few of the regression tests failed. Is this normal,
> or should I
> submit a bug report?

This is normal, if you look at the results in regression.diff there will be
no important differencies.

>
> I'm still interested to know if anyone Out There has a working Win32
> Postgres database doing real work.

I personally don't use it for a production system, I am waiting for outer
joins. Than I will test it in my software.

Dan