From: Ernest Ellingson <erne@...>
Date: 2005-01-09T04:51:26+09:00
Subject: Re: postgres-pr

Michael Neumann wrote:
> Ernest Ellingson wrote:
> 
>> Michael Neumann wrote:
>>
>>> Ernest Ellingson wrote:
>>>
>>>> Thanks Michael
>>>> I'm still not sure about Pg.rb.  Where is it?  It's not in the DBI 
>>>> or the DBD directory.  Does it exist?  How do I get it?
>>>
>>>
>>>
>>>
>>> Ah, okay. I think it's not bundled with the Windows-Installer, as you 
>>> need the postgres.so or postgres-pr libraries. And postgres.so is/was 
>>> not available for Windows AFAIK.
>>>
>>> You can download it from here:
>>>
>>> http://ntecs.de/viewcvs/viewcvs/dbd-pg/trunk/src/?root=ruby-dbi
>>>
>>> Just put the Pg.rb into the DBD/Pg directory (create the Pg directory).
>>>
>>>> As for postgres-pr I tried the test_connection.rb code changed the 
>>>> uri to 'tcp://ipaddress:5432' and the other parameters to 
>>>> appropriate values.  Apparently a connection is being made but the 
>>>> same blow up in page files occurs 1.25 GB's.  The program finally 
>>>> errors out with this trace. I should mention the database is on 
>>>> another machine running RH linux 9 and postgresql 7.4.
>>>
>>>
>>>
>>>
>>> Hm, very strange. Does using TCPSockets generally work on your 
>>> platform without problems?
>>>
>>>> c:/ruby/lib/ruby/gems/1.8/gems/postgres-pr-0.3.3/lib/buffer.rb:64:in 
>>>> `read': failed to allocate memory (NoMemoryError)
>>>
>>>
>>>
>>>
>>> I should include some debug code. It might be, that there's a 
>>> negative buffer-length which gets interpreted as an unsigned value... 
>>> or something like this. You're on a 32-bit machine, right?
>>>
>>> Regards,
>>>
>>>   Michael
>>>
>>>
>> I haven't had a problem with smtp or gets or posts with httpAccess2 so 
>> I believe that TCPsockets are working fine.  I am running a 32 bit 
>> machine
>> WindowsXP svc Pack 2.
> 
> 
> Could you patch message.rb of postgres-pr with the following patch, and 
> tell me what you get on STDOUT:
> 
> Index: message.rb
> ===================================================================
> --- message.rb  (revision 475)
> +++ message.rb  (working copy)
> @@ -32,6 +32,8 @@
>      type = stream.readbytes(1).unpack('C').first unless startup
>      length = stream.readbytes(4).unpack('N').first  # FIXME: length 
> should be signed, not unsigned
> 
> +    p length
> +
>      raise ParseError unless length >= 4
> 
>      # initialize buffer
> 
> Regards,
> 
>   Michael
> 
> 
Hi Michael:
	The p length statement yields

	1178686529

Which is where the page file blow up happens.

I put a p type line in too.
the type is
	69

Ernie