Activity (RSS)

I am running this on Debian Squeeze.
It's possibly related to the way the memory is allocated then released.

I don't have other test case than the "testmem.py" ...

You're getting it wrong. Let me try to explain. Under Linux (and Solaris I guess) you can do dynamic or static linking. Static linking means including a version...

Thanks for your help here.
I've tried with static_libpq=1 in setup.cfg; after a successful build, output of ldd is:

$ ldd build/lib.solaris-2.11-i86pc-2.6/psyc...

setup.py is linking in libpq. What you want is to embed a statically linked version of libpq into the _psycopg.so module. You can try to pass the --static-libpq...

Federico, Tks for that response. As I mentioned, I am able to direct psycopg2 to libpq at runtime with the LD_LIBRARY_PATH (meant to be only a 'technology demon...

Build-time linking and run-time linking are different things. This is how run-time (dynamic) linking works: you need to put the libpq library into one of the de...

This is better:

if (stack_index == 0)
    return -1;
array = stack[--stack_index];

and should fix the warning.

The warning is correct but using <= instead of < would break array parsing. That test is a guard to avoid the (impossible?) case of receiving more ASCAN_END tha...

Submitted by: mjl@laubach.at

The clang compiler complains that:

In file included from psycopg/typecast.c:179:
    ./psycopg/typecast_array.c:227:31: warning: c...

Got no reply back. However I've merged the patch and added test and docs myself.

I've made scrollable three-states: None (default) specifies no scroll option. T...

Fixed in my devel. Thank you for the report.

I'm putting this bug on hold: I haven't got enough info to fix it. Please provide a test case if you care.

So it's no link in psycopg, just the way the python gc works, right?

Hello Joshua,

register_composite works for me.

In psql:

psql (9.0.3, server 9.0.5)
Type "help" for help.

piro=> create type foo as  (a integer, b integer);
CR...

I am running psycopg2 2.4.5 on python 3.2.2.

I found error in psycopg2.extras.register_composite which is probably caused because of usage of old formatting of ...

I changed the test_fetch() function to release the memory.
For some reason, "del big_var" is not enough. A "gc.collect()" call is required.

def test_fetch(curs...

This is the test file attached.

And the output:

$ python testmem.py psycopg2 version 2.4.5 (dt dec pq3 ext)
Record size: 1024 B; Memory used: 8.8 MB
Record size...

Submitted by: flox

When fetching a named cursor with large records > 16340 bytes, the memory is not released.

Does it hang forever or just takes longer to complete?

Can you provide a self-contained test case for the problem?

Thank you.

Submitted by: Jeff Shanab

Traceback (most recent call last):
  File "C:\Program Files (x86)\MVSArchiveServerD_x86-1.2.15\scripts\InstallDb.py", line 504, in <m...

Fixed, thank you.

Submitted by: niels.olson@gmail.com

In Python Types Adaptation, the word standards should be singular.

http://initd.org/psycopg/docs/usage.html#python-types-ada...

Daniele,

The original bug isn't mine, I'm just adding to it because it seems to be the exact same symptoms. It's Anoop's bug originally.

I am using 2.4.5.

We'll...

Josh, another possible source of help: try compiling and running psycopg with debug info (setting PSYCOPG_DEBUG in setup.cfg and running the program with the PS...

Josh, are you using the latest, 2.4.5? There have been a few multithread bug fixed in that version.

Anoop, from your tag I see you are using 2.4.1: can you try ...

I'm seeing the exact same problem in a multithreaded Python application. It makes and drops connections rapidfire for a few minutes, then hangs. GDB indicates t...

The problem is in psyco_clear_result_blocking: it is called because the network problem has been detected. That function was supposed to be called upon an error...

In the meantime would it be possible to change the documentation to make this plain.

Yes, several psycopg parts don't work correctly with RealDict cursors, because they are not dbapi-compatible. Supplying the OID is the designed workaround.

OTOH...

Submitted by: Adrian Klaver

Seems the register_hstore only works with connections that use cursors that return an indexed tuple. When using the RealDictConnecti...

Submitted by: anoop

I have a program which keeps a persistent connection to a postgresql server. Suppose the connection is blocked due to network issues(can be ...

Submitted by: anoop

We have a python program which uses psycopg2 to make a connection every 2 seconds to postgresql-9.0 and batch insert/update records. The con...

Adding the following symlinks to /usr/lib fixed this problem for me on Mac OS X 10.6.8 with PostgreSQL 9.0.7 (EnterpriseDB installer) and psycopg2 2.4.5:

libssl...

Hello Chad,

this is a problem with your build chain. None of us uses OSX for development so we are not able to debug it.

If you cannot debug it yourself, please...

Submitted by: chad@zetaweb.com

I can build psycopg2 but when I import it I get:

Traceback (most recent call last):
File "", line 1, in File "/Users/whit537/pers...

This bug was terrible, I just wonder why it hasn't bit us before. So I'm for releasing soon, even if its longevity tells it's not hot urgent.

I try and run the ...

Thanks,

I would just like to know if this is going to mean a new release soon, or if I should build the module from a patched fork for the time being.

Manu

No problem :-)
Thanks for fixing setup.cfg

Committed in my devel.

Thank you Steven, sorry for asking you a patch that we ended up not merging.

That's fine with me.

Ok with something like that?

diff --git a/setup.cfg b/setup.cfg
index db4e771..ce17624 100644
--- a/setup.cfg
+++ b/setup.cfg
@@ -22,26 +22,11 @@ have_ssl=0
 #...

Yes, we should remove it before next release and replace it with an explanation text.

Definitely we should sort out the inconsistency. Fog, shall we remove the *_dirs from setup.cfg?

Fixed in my devel. Thank you again, Manu.

I'd avoid switching to malloc() because, at least in theory, PyMem_Malloc tells the interpreter how much memory pressure we're putting on it and that's a Good T...

At a first glance that seems the only place where we call psycopg_escape_string without GIL, but couldn't tell for other PyMem_Malloc.

I came to the same conc...

I think you are right, thank you for the report.

At a first glance that seems the only place where we call psycopg_escape_string without GIL, but couldn't tell ...