Mailman 3 June 2008 - pypy-dev

channel logs
by Carl Friedrich Bolz 01 Jul '08

01 Jul '08

Hi all, just wanted to note that the braintone logs seem to work at the moment, so you can read the as a stopgap measure here: http://merlinux.de/braintone/chat.freenode.net/pypy/last/50/ The format is not so nice, so I would actually really like pybot back :-). Cheers, Carl Friedrich

3 2

pyrepl: cursor blinks
by Armin Rigo 30 Jun '08

30 Jun '08

Hi all, When I'm running an interactive pypy-c over an ssh connection I notice that the cursor blinks a bit when I type (it can remain invisible for a fraction of second). Normal readline-style editing doesn't give this result even over laggy connections. I traced it to pypy/lib/pyrepl/unix_console.py lines 95 and 97, which are the control code for hiding and re-showing the cursor, called all the time when refreshing the display. I really don't like this, but of course it is not a concern for people whose terminal's cursor is blinking all the time anyway. Anyway, if I remove these control codes, I get quite a fine experience again. However, I'm sure this is highly dependent on the terminal and I fear that some terminals might display the cursor moving quickly around while refreshing is taking place. If anyone knows of a better solution to avoid both issues, please do tell :-) Otherwise, if you feel like it, could you please try the attached patch in your favourite terminal? No need to re-translate a pypy-c. Tell me what your experience is, running an interactive pypy-c over an ssh connection (preferably laggy). A bientot, Armin.

2 1

I reinstated the nightly run of cpython tests against pypy-c
by Samuele Pedroni 29 Jun '08

29 Jun '08

I switched to only test these combinations: 1) hybrid gc thread old-style allworkingmodules the result of a try run for this one can be seen at the usual place: http://www2.openend.se/~pedronis/pypy-c-test/allworkingmodules/summary.html http://www2.openend.se/~pedronis/pypy-c-test/allworkingmodules/56156/index.… although just the % doesn't tell much it seems we had some regressions in some fundamental tests, like test_builtin also our modified test_descr segfaults with this combination/these days it seems 2) stackless old-style allworkingmodules

2 2

Dotviewer capabilities
by Dinu Gherman 26 Jun '08

26 Jun '08

Hi all, while preparing a talk for EuroPython.org and chatting with Christian Tismer I remembered that PyPy had a nice PyGame GUI for browsing GraphViz .dot files. So I wanted to play with it and made a new SVN checkout (very funky Mandelbrot sets during compilation, BTW... ;-) Then it occured to me that I don't need PyPy to run dotviewer (great, because it's still compiling). So I throwed a few of my .dot files at dotviewer and got slightly surprised to see a few ra- ther basic things not quite working as I'd expect (like rec- tangular and record-like node shapes. I've created the follow- ing sample files to illustrate my findings: http://python.net/~gherman/tmp/sample.dot http://python.net/~gherman/tmp/sample.pdf http://python.net/~gherman/tmp/sample.sample-dotviewer.jpg I'm running PyGame 1.8.0rc5 on Mac OS X 10.5.3, BTW. So, am I running dotviewer in some wrong fashion? Do I need to run it on a different PyGame version or another OS? Or is there some documentation for docviewer and its capabilities, that I haven't found yet? In fact, what I hoped to find in dotviewer is a browser for dot files in which I can fold individual parts of a graph in and out (while GraphViz is, perhaps, recalculating the layout). If dotviewer cannot do that, maybe somebody knows of another browser which can? Thanks and regards, Dinu

5 5

changes to pysqlite/ctypes
by Maciej Fijalkowski 23 Jun '08

23 Jun '08

We implemented few changes, also new features to sqlite-ctypes. Most notably: * Implement create_function without py_object, using closures * Various bugfixes * Accept subclasses of string/unicode * Added support for readahead logic * Thread safety All with tests, or reenabling of tests. However, when I try to use your supplied info how to commit there, it says "authorization failed". Would you rather prefer it in patches or commited to the repo? Cheers, fijal

2 3

Benchmarking django admin app
by Maciej Fijalkowski 22 Jun '08

22 Jun '08

I did some benchmarks for django admin app. It was fairly hard since django is good at eating exceptions (there are like 3 levels that can eat your SystemExit) which makes it hard to correctly exit application. Anyway, it seems that our ctypes is huge problem there, calling of ctypes functions eats 20% of time overall (which is fairly a lot for 1-value table). C-level benchmark does not show anything specific. Overall it's about 3x slower than cpython, which is fairly bad (it uses the same ctypes-based pysqlite). I think we should repeat this exercise after faster ctypes and see how this goes. I don't have any reasonable benchmark to try without DB, since default webpage takes 4ms to create (12 on pypy-c), but it can be some small issue that makes huge difference. (And we don't support builtin benchmarking for pypy) Cheers, fijal

2 2

Default options for pypy
by Maciej Fijalkowski 20 Jun '08

20 Jun '08

Can I make --gc=hybrid --thread --faassen --allworkingmodules --oldstyle default options for pypy? People get easily confused.

1 0

Re: [pypy-dev] From Brett in infrastructure
by Laura Creighton 20 Jun '08

20 Jun '08

In a message of Fri, 20 Jun 2008 08:54:24 +0200, Armin Rigo writes: >Hi Laura, > >On Fri, Jun 20, 2008 at 05:17:10AM +0200, Laura Creighton wrote: >> Do we have a tool that does this? > >Not that I know of. I don't even really know the svnmerge.py he is >talking about and we haven't used it in the pypy repository as far as I >am aware. > > >Armin Ah, well I was confused. But it turns out so was he, and the tool he wants already comes with svn. Laura

1 0

From Brett in infrastructure
by Laura Creighton 20 Jun '08

20 Jun '08

Do we have a tool that does this? Laura ------- Forwarded Message Return-Path: infrastructure-bounces+lac=openend.se(a)python.org Delivery-Date: Fri Jun 20 04:41:53 2008 Return-Path: <infrastructure-bounces+lac=openend.se(a)python.org> From: "Brett Cannon" <brett(a)python.org> To: infrastructure(a)python.org Subject: [Infrastructure] How do we move over to svn 1.5 once 2.6/3.0 final are out? Does anyone happen to know if there is any way to use ``svn merge`` with a branch that has been managed with svnmerge.py initially? I would love to get the built-in support once we are out of beta, but I am not sure how to make sure we don't lose all the info we currently have going in branches such as py3k. - -Brett

2 1

relative occurency of opcodes
by Maciej Fijalkowski 18 Jun '08

18 Jun '08

I just performed simple statistics on hacked pypy-c. It seems that some opcodes are very often repeated by something else. It might make sense to create some super opcodes which combine those two. Here are statistics for translating gcbench: http://codespeak.net/svn/user/fijal/opcodes.txt

2 1