[01:03] yuuh (dkfynkv@82.140.29.224) left irc: Read error: 131 (Connection reset by peer)
[01:07] dialtone (~dialtone@host111-56.pool80117.interbusiness.it) left irc: "This computer has gone to sleep"
[01:15] dialtone (~dialtone@host111-56.pool80117.interbusiness.it) joined #pypy.
----- silence for 3 hr and 58 minutes ----- [05:13] tewk (tewk@192.41.88.101) joined #pypy.
[05:26] esteban (~esteban@DWM-53-230.go.retevision.es) joined #pypy.
[05:37] esteban (esteban@DWM-53-230.go.retevision.es) left #pypy.
----- silence for 42 minutes ----- [06:19] idnar (mithrandi@idnar.user) got netsplit.
[06:19] hpk (~hpk@merlinux.de) got netsplit.
[06:19] etrepum (bob@ayunami.redivi.com) got netsplit.
[06:19] tic (~vision@s-213-115-110-186.sme.bredbandsbolaget.se) got netsplit.
[06:20] idnar (mithrandi@idnar.user) returned to #pypy.
[06:20] hpk (~hpk@merlinux.de) returned to #pypy.
[06:20] etrepum (bob@ayunami.redivi.com) returned to #pypy.
[06:20] tic (~vision@s-213-115-110-186.sme.bredbandsbolaget.se) returned to #pypy.
----- silence for 53 minutes ----- [07:13] idnar (mithrandi@idnar.user) left irc: Read error: 110 (Connection timed out)
----- silence for 20 minutes ----- [07:33] arre (ac@ratthing-b3fa.strakt.com) joined #pypy.
----- silence for 1 hr and 24 minutes ----- [08:57] idnar (mithy@idnar.user) joined #pypy.
----- silence for 1 hr and 15 minutes ----- [10:12] sanxiyn (tinuviel@211.104.100.177) joined #pypy.
----- silence for 21 minutes ----- [10:33] sanxiyn (tinuviel@211.104.100.177) left irc: "Bye"
----- silence for 33 minutes ----- [11:06] ludal (~ludal@logilab.net2.nerim.net) joined #pypy.
----- silence for 33 minutes ----- [11:39] <hpk> ludal: hi ludovic
[11:40] <hpk> test_samples.py keeps failing, does it work for you?
----- silence for 30 minutes ----- [12:10] lac (~lac@lac.silver.supporter.pdpc) joined #pypy.
[12:10] <lac> hi all
[12:13] arigo (~arigo@ratthing-b407.strakt.com) joined #pypy.
----- silence for 20 minutes ----- [12:33] <lac> hi armin
[12:33] <lac> where are you<?
[12:33] <lac> strakt it seems :-)
[12:36] <ludal> hi hpk
[12:37] <ludal> test_samples.py has one test that fails
[12:38] <lac> hi ludal
[12:38] <ludal> it's kind of low priority fix : the parser sometimes provide one too many newlines at the end of a file (when there are comments)
[12:38] <ludal> hi
[12:38] <lac> ah
[12:39] hpk (~hpk@merlinux.de) left irc: Read error: 113 (No route to host)
[12:39] <lac> its ok here on the ws, but
[12:39] <lac> you are correct for my laptop
[12:39] <lac> wonder what makes the difrerence+
[12:40] <lac> ?
[12:40] <lac> not +
[12:40] <lac> new keyboard
[12:43] <arigo> hi
[12:43] <arigo> yes I'm at strakt, at your desktop to be more precise :-)
[12:44] <arigo> ludal: actually, depending on the machine, I see anywhere between 1 and 4 failures
[12:46] <ludal> arigo: could you email me the traceback
[12:46] <arigo> ok, I'll try to find a 4-failures machine then :-)
[12:46] <ludal> is this with python2.3 or 2.4
[12:46] <lac> arigo: and pedronis: TONIGHT is the deplomna consert of the new music masters.
[12:46] <arigo> always 2.3
[12:46] <ludal> ok
[12:47] <lac> this is free.
[12:47] <arigo> lac: ah ha
[12:47] <ludal> well even 2 is ok; I only see one
[12:47] <lac> want to come?
[12:47] <arigo> yes, sure
[12:47] <arigo> Samuele's not around for now, I can't ask him
[12:48] <arigo> ludal: if you log on codespeak.net and try there, you'll probably see three failures
[12:48] <arigo> ludal: I can still send you the tracebacks if you prefer
[12:48] <ludal> ok
[12:49] <ludal> hold on...
[12:50] <ludal> any checked out tree on codespeak I can use ?
[12:50] <arigo> check out yours, or use mine (read-only, though) in ~arigo/pypysrc/pypy
[12:51] <arigo> checking out on codespeak is cheap :-)
[12:51] <ludal> I just want a quick check whether I can get those traceback
[12:51] <ludal> ok
[12:51] <lac> i have one in ~lac but its a month out of date ...
[12:53] <ludal> ok got them
[12:53] <ludal> no need to email me
[12:54] <arigo> ok
[12:54] Action: ludal running off to catch the others before there's nothing left to eat
[12:59] <lac> arigo its free entrance for a concert at 19.30
[13:00] <lac> Diplomkonsert - Fri entré
[13:00] <lac> Rouse - Flöjtkonsert ffg
[13:00] <lac> Sunde - Comatose, uruppförande
[13:00] <lac> Bartók - Violinkonsert nr 1
[13:00] <lac>
[13:00] <lac> Göteborgs Symfoniker
[13:00] <lac> Dirigent: Johannes Gustavsson
[13:00] <lac>
[13:00] <lac> Medverkande diplomander:
[13:00] <lac> Cibran Sierra-Vazques, violin
[13:01] <lac> Jill Widén, flöjt
[13:01] <lac> Knut Olaf Sunde, komposition
[13:01] <lac>
[13:01] <lac> when and where to meet+
[13:01] <lac> ER meet?
[13:01] <arigo> I have a g/+/? filter for you :-)
[13:01] <lac> (NEW KEYBOARD)
[13:02] <lac> :-)
[13:02] <arigo> where is that? same as last time?
[13:02] <lac> yes. so something like 3 block of my house
[13:02] <arigo> I see
[13:03] <lac> meet before, after, dinner ...
[13:03] <lac> eat out eat in at my place ... i need to buy food if we are all to eat here.
[13:04] <lac> is iko around?
[13:04] <arigo> probably, though I'm not sure I remember who is iko at this point
[13:04] <lac> (anders hammarquist)
[13:04] hpk (~hpk@merlinux.de) joined #pypy.
[13:05] <lac> hi holger. long time no chat
[13:05] <lac> nice to see merlinux.de is alive
[13:07] <hpk> lac: hi Laura!
[13:07] <hpk> indeed long time not heard from each other
[13:07] <hpk> hope it is going well for you
[13:08] <lac> seems better now.
[13:08] <lac> and for you? and your mother?
[13:09] <hpk> lac: all kind of ok, not particularly good, though (let's not talk about this here)
[13:14] Action: hpk -> lunch
[13:15] <lac> q
[13:22] yuuh (oqeesx@82.140.29.187) joined #pypy.
[13:23] <lac> http://www.ans.org/pi/np/oklo/
[13:31] cfbolz (~cfbolz@edv01.physi.uni-heidelberg.de) joined #pypy.
[13:31] <cfbolz> hi all!
[13:31] <arigo> hi!
[13:32] <lac> hi hi
[13:32] <ludal> hpk, armin : any idea why on codespeak: import symbol; symbol.encoding_decl == 324
[13:32] <ludal> and here (debian python 2.3.5) symbol.encoding_decl == 323
[13:33] <ludal> same python version, probably compiled with different flags on gentoo?
[13:34] <lac> smart idea.
[13:35] <arigo> no clue indeed
[13:35] <lac> but since I DO not know who makes symbol.encoding
[13:35] <ludal> looks like symbol.py is generated at compile time
[13:35] <lac> no clue from me.'
[13:35] <ludal> 323/324 probably map to DFA state number of the grammar
[13:36] <ludal> but I can't see why there would be a different parser generated from the same version of python...
[13:36] <arigo> maybe the numbers are generated in a non-reproducible way (e.g. depending on dict key order somewhere) ??
[13:37] <ludal> short answer: on debian 2.3.5 there is an iexec_stmt that's inserted in the list of grammar symbol
[13:37] <arigo> argh argh argh
[13:37] <ludal> all other symbols map exactly
[13:38] <arigo> that's Holger's extension
[13:38] <ludal> and I was wrong the iexec is indeed on codespeak
[13:38] <arigo> it must be by accident that the globally installed codespeak Python version is the one Holger hacked
[13:39] <arigo> yes, it has a syntax extension:
[13:39] <arigo> try to type
[13:39] <arigo> <5>:
[13:39] <arigo> 6
[13:39] <arigo> this is valid syntax for codespeak's python :-)
[13:39] <cfbolz> what does it do?
[13:40] <arigo> well it's something similar to the proposed 'block' statement discussed on python-dev these days
[13:40] <ludal> anyway the test failures can easily be fixed by using symbol.py because what's going wrong is a hack for handling encoding declaration and the value has been hardcoded instead of the symbol...
[13:40] <arigo> makes sense
[13:40] <cfbolz> arigo: ok. I didn't follow that.
[13:40] <arigo> good of you
[13:40] <cfbolz> that bad?
[13:41] <arigo> there is almost one mail per minute on python-dev, these days
[13:41] <ludal> it's just verbose on pytohn-dev
[13:41] <arigo> Holger's extension is:
[13:41] <arigo> <expr>:
[13:41] <arigo> block
[13:42] <arigo> expr is evaluated and then the block is executed with the 'expr' object in control of things like raised exceptions
[13:42] <arigo> with call-backs like __enter__(), __leave__(), __except__(type,value,traceback)
[13:42] <ludal> in control meaning?
[13:42] <ludal> ok
[13:42] <cfbolz> cool
[13:43] <arigo> it is also a dynamic scoping feature,
[13:43] <arigo> because there is a simple way to look up the current stack of handlers and ask e.g. for one which would be an instance of some specific class
[13:44] <arigo> see e.g. http://codespeak.net/svn/pypy/www/pypywww/render/doc.py
[13:44] <arigo> where it's used to build the dynamic web pages of PyPy.
[13:46] <cfbolz> looks very XML-ish
[13:49] <arigo> it was meant this way, to look like html tags
[13:52] <cfbolz> can I look at the implementation?
[13:52] <arigo> http://codespeak.net/svn/xpython/trunk/dist/src/
[13:53] <cfbolz> thanks
[13:58] <lac> owi owi owi
----- silence for 16 minutes ----- [14:14] <cfbolz> test_profile fails due to some float formatting problem: '%8.3f' % 0.04987 is ' 0.050' in CPython but ' 0.049' in PyPy
[14:16] <arigo> aaah, thanks for the hint
[14:16] <arigo> we thought the problem was about off-by-one counts
[14:16] <cfbolz> it's not those exact numbers, though
[14:17] <cfbolz> I just ripped out the formatting functions of pstats and tried
[14:19] <hpk> arigo: indeed, on codespeak x-python is globally installed with gentoo to get all the dependencies like svn-bindings etc.pp., i should probably have a plain python2.3.5 for pypy-testing somewhere
[14:20] <arigo> :-)
[14:20] <hpk> the patch is amazingly stable since 2.2 up to 2.4 (no modifications basically)
[14:21] <arigo> :-)
[14:22] <hpk> regarding test_exceptions.py i pushed it to the point where it only has exception-representation problems (output-test)
[14:22] pedronis (pedronis@ratthing-b246.strakt.com) joined #pypy.
[14:22] <hpk> but i got a bit confused regarding how to actually fix it by a small change somewhere
[14:27] <cfbolz> I'm going home. See you all.
[14:29] <hpk> see you
[14:30] cfbolz (~cfbolz@edv01.physi.uni-heidelberg.de) left irc: "Leaving"
[14:36] <lac> hi samuele
[14:38] <pedronis> lac: hi
[14:44] fredrik (fredrik@c83-248-135-181.bredband.comhem.se) joined #pypy.
[14:45] <hpk> arigo, pedronis: i am thinking of renaming the coding-style.txt to coding-guide.txt and adding the wrapping-chapter from objspace.txt
[14:46] <hpk> then it could also nicely contain a chapter on how to implement modules
[14:47] <pedronis> I see. Armin right now is on a grabbing food mission
[14:47] <hpk> 9 minutes left :-)
[14:49] <arigo> hpk: back
[14:49] <arigo> coding-guide makes sense
[14:55] stakkars (~tismer@rosine165.inf.fu-berlin.de) joined #pypy.
[14:56] <lac> HEJ christian!
[14:57] <stakkars> hi Laura :)
[14:57] <lac> long time no write.
[14:57] <stakkars> yes. I was very busy
[14:57] <lac> me as well
[14:58] <hpk> arigo, pedronis: it's time
[15:00] <lac> tech meeting?
[15:00] <stakkars> yup
[15:00] <lac> if samuele sent armin to forage after his second day here, then he might be late ....
[15:01] <stakkars> he is there
[15:01] <pedronis> lac: no, it was Armin in autonomous mode
[15:02] <stakkars> herbaceous plant material eaten by grazing animals?
[15:03] <lac> I was going to forage myself. I am correct in assuming that this meeting is not for me?
[15:04] <arigo> lac: yes, it's the technical board offically.
[15:04] <stakkars> if you like to listen, there is nothing secret
[15:04] <pedronis> lac: no, it's the technical board
[15:05] <lac> I always like listening, but I am also hungry.
[15:05] <lac> would not leave if this was a meeting i was supposed to attend but forgot . :-)
[15:06] <lac> mosty i would like to /join where this is and just be able to read it.
[15:08] <arigo> lac: #pypy-tb
----- silence for 1 hr and 29 minutes ----- [16:37] Nick change: lac -> lac-afk
[16:52] Nick change: lac-afk -> lac
----- silence for 24 minutes ----- [17:16] idnar (mithy@idnar.user) left irc: "goats"
[17:25] ludal (ludal@logilab.net2.nerim.net) left #pypy.
[17:31] adim (~adim@logilab.net2.nerim.net) joined #pypy.
[17:32] adim (adim@logilab.net2.nerim.net) left #pypy.
[17:43] <arigo> lac: can you suggest a plan for tonight
[17:43] <arigo> ?
[17:52] <stakkars> arigo: a question about buildcache2:
[17:54] <arigo> stakkars: I'm working on deprecating it... it would be nice not to use it if we are doing annotations
[17:55] <stakkars> why is it needed? rpystone works fine without it,so I guess it is to initialize the object space.
[17:55] <stakkars> but I thought by poking at the objects needed, this would be implicitly?
[18:01] <stakkars> erhm, was there a checkin since yesterday that might crash thegoals?
[18:01] <arigo> it's needed to initalize the caches, because the annotator isn't smart enough so far to do it "just-in-time" when it sees the caches in question
[18:01] <arigo> stakkars: yes
[18:01] <stakkars> I get concrete call with a non-constant arg %r" % (s,)
[18:02] <arigo> Samuele and me are poking with this code at the moment
[18:03] <stakkars> so I might revert to an earlier version. Should I keepmy hands off of the annotator?
[18:03] <arigo> it's more involved that I thought at first, that's why I checked in stuff and then realized it's not so simple
[18:03] <arigo> stakkars: no, go ahead, the annotator is only changing in a very local place in bookkeeper.py
[18:03] <stakkars> what do I need to get it running, again?
[18:04] <stakkars> this is the part of the code that I still don't understand (but it's getting less)
[18:04] <arigo> a revision from 12 o'clock would do, I guess
[18:04] <stakkars> thanks
[18:13] arre (ac@ratthing-b3fa.strakt.com) left irc: Remote closed the connection
----- silence for 27 minutes ----- [18:40] arigo (~arigo@ratthing-b407.strakt.com) left irc: "[x]chat"
[18:40] pedronis (pedronis@ratthing-b246.strakt.com) left irc: "Chatzilla 0.9.67 [Firefox 1.0.2/20050325]"
[18:55] idnar (~mithrandi@idnar.user) joined #pypy.
[18:55] <stakkars> I still have no clue why targetpypy1 with "div" works fine,
[18:56] <stakkars> but targetpypy1 with "floordiv" generates everything.
[18:56] <stakkars> especially because div is implementedon top of floordiv.
[18:56] <stakkars> Can it be that this depends on the order of flow-graphing?
[18:57] idnar (~mithrandi@idnar.user) left irc: Client Quit
[18:57] <stakkars> But I stilldon't get it. floordiv happens to trigger a descroperation, and that's it - boom
[19:00] <stakkars> How can I debug this?
[19:03] idnar (mithrandi@idnar.user) joined #pypy.
[19:05] cfbolz (~cfbolz@hdlb-d9b946a9.pool.mediaWays.net) joined #pypy.
[19:06] <cfbolz> hi!
[19:06] <stakkars> hi
[19:06] <cfbolz> I just read the technical board summary.
[19:06] <cfbolz> I will probably be avaiable for polishing between the 15th and 20th
[19:08] <stakkars> great
[19:08] <cfbolz> :-)
[19:19] tic (~vision@s-213-115-110-186.sme.bredbandsbolaget.se) left irc: Read error: 54 (Connection reset by peer)
[19:20] tic (~vision@s-213-115-110-186.sme.bredbandsbolaget.se) joined #pypy.
[19:28] <cfbolz> bye
[19:29] cfbolz (~cfbolz@hdlb-d9b946a9.pool.mediaWays.net) left irc: "Verlassend"
----- silence for 1 hr and 24 minutes ----- [20:53] aleale (~root@cpe.atm0-0-0-129140.0x3ef2fa3a.bynxx3.customer.tele.dk) joined #pypy.
[20:54] <aleale> hi
[20:54] <aleale> Any news on the mini sprint?
[20:55] <tic> having a sprint soon?
[20:56] <aleale> There was talk about a mini sprint the 15-20th of may
[20:58] <aleale> hpk: I am having trouble running pytest/conftest.py
[20:59] <stakkars> aleale: cancelled, we work via irc and pypy-dev
[21:00] <aleale> hpk: It barfs at conftest line 880 "os.WIFEXITED" with an attributeerror
[21:00] <aleale> Is os.WIFEXITED only on linux ? I am running win2k
[21:01] <hpk> aleale: hi anders
[21:01] <hpk> when armin and wrote the that part of the code we didn't know if it would work on win32
[21:02] <hpk> the question is how to get at a meaningful exit status of 'os.system()' call
[21:02] <stakkars> fine to have someone else :)
[21:02] <hpk> stakkars, aleale: does anyone of you know how to do that?
[21:02] <aleale> stakkars: Ok, thankyou well I was hoping to join you there
[21:02] <aleale> No but I will try to look into it
[21:03] <hpk> well the whole getstatusouterr() function is somewhat unix-specific
[21:03] <stakkars> os.system:
[21:03] <hpk> but we wanted to get going and so far noone complained :-)
[21:03] <stakkars> "meaningful" is just 0 for no error
[21:03] <hpk> anything other than 0 means an error?
[21:03] <aleale> while we are at windows specific topics. The codecs module has som functions concerning MBCS (multibyte character sets)
[21:03] <hpk> aleale: wait a moment
[21:04] <aleale> Ok
[21:04] <hpk> stakkars: maybe you could look into lib-python/conftest.py 'def getstatusouterr' to see how we could adapt that for win32?
[21:05] Action: hpk hopes this is possible on win32 with reasonable effort
[21:07] <hpk> aleale: regarding codecs: i am probably not the right person to talk to, especially not this evening, i am busy with other stuff right now
[21:07] <hpk> do you actually read pypy-funding?
[21:09] <aleale> I forgot my password so I cant access it, and since I am not a partner it didnt feel right to ask for a new one - so no I dont read pypy-funding
[21:10] <hpk> aleale: not a partner? but you are working for dfki, no?
[21:10] <aleale> yes but DFKI is the partner. I am only an employee.
[21:11] <hpk> aleale: well, everyone involved with the EU project should be on pypy-funding
[21:11] <aleale> But I would like to be able to read pypy-funding
[21:12] <aleale> but then I need to ask you for a new password
[21:12] <tic> where's the sprint going to be?
[21:12] <tic> (if)
[21:12] <hpk> tic: here :-)
[21:12] <hpk> in fact we are going to do a concentrated "release-polisihing" irc-coordinated hacking session
[21:13] <hpk> aleale: your email?
[21:14] <tic> in deutschland indeed.
[21:14] <hpk> tic: no, here on IRC
[21:14] <tic> Oh.
[21:14] <hpk> we are going to meet before EuroPython and after EuroPython as the next sprints
[21:14] <tic> mhm.
[21:14] <tic> where will you do the sprint after and before EP?
[21:14] <hpk> goetheborg
[21:15] <tic> ah, cool.
[21:15] <tic> maybe I'll have a look at the development process :)
[21:15] <aleale> hpk: I'll mail it to you privately. ok?
[21:16] <hpk> aleale: i opened a private channel ...
[21:19] <aleale> hpk: I sorry, a bit paranoid here it is : serendipity-soft@get2net.dk. I dont know how to access a private channel
[21:19] <aleale> serendipity-soft@get2net.dk
[21:21] <hpk> aleale: done
[21:22] <hpk> aleale: you should see the private channel somewhere with your IRC client
[21:29] <aleale> I see it but dont know to access it. Tried /join -invite
[21:30] <aleale> Well I have to go. What are the best times to catch you here (all the time ?)
[21:31] <hpk> aleale: usually around the clock, yes, on the day there are more people available usually but it all depends
[21:31] <aleale> Ok se you tomorrow
[21:31] aleale (~root@cpe.atm0-0-0-129140.0x3ef2fa3a.bynxx3.customer.tele.dk) left irc: "What's a script?"
[21:31] <hpk> see you
[21:35] <stakkars> hpk: I tried it since you asked, but with no good result at all.
[21:36] <stakkars> even the example from
[21:36] <stakkars> http://mail.python.org/pipermail/python-list/2003-December/198917.html
[21:36] <stakkars> does not give me return codes.
[21:36] <hpk> damn it
[21:36] <stakkars> maybe it is different between 2K and XP
[21:40] <hpk> i guess the subprocess module could do what we want but it is included only on python 2.4 i think
[21:40] <stakkars> the only thing I can generate is an "1" if python exists due a syntaxerror, for instance#
[21:41] <stakkars> maybe
[21:41] <stakkars> things are broken, I know I have used this before.
[21:45] <hpk> have you tried subprocess?
[21:45] <hpk> i think it should do what we want
[21:45] <hpk> it reuqires Mark's win32 extensions though (which should be reasonable)
[21:48] <stakkars> I'm on py 2.3.3
[21:49] <stakkars> ah, ok, win32process will work, of course.
[21:49] <stakkars> I thoguht you wantedsomething like a compromize
[21:49] <stakkars> do you have the *slightest* idea what the difference between "div" and "floordiv" is, in terms of how they
[21:50] <stakkars> are implemented or registered or something?
[21:50] <hpk> not without investigating myself
[21:50] <stakkars> div actually calls floordiv
[21:50] <stakkars> I can nicely run targetpypy against div. But notfloordiv.Samestory with rshift andlshift.
[21:51] <stakkars> it drives me crazy,somehow a descroperation is generated, and that's it:whole interpreter.
[21:51] <stakkars> I just need a clue how to debug that!
[21:53] <hpk> if you follow the long of the annotator you can't determine what drags in the descriptor?
[21:54] <hpk> I am not really suprised, btw, that you can easily drag in the whole interpreter
[21:54] <stakkars> no, I'm trying since hours.
[21:54] <stakkars> yes, but div is a caller of floordiv :-(
[21:54] <stakkars> even if I pull them both in, first div, then floordiv, no change.
[21:54] <stakkars> What makes me curious:
[21:55] <stakkars> when I use div,it even works without filling the cache!
[21:55] <stakkars> in the case of floordiv, I get a huge complaint.
[21:55] <hpk> are you talking about space.div?
[21:56] <stakkars> I think so. It's the multimethod.
[21:56] <hpk> did you know that you can do the folliwing in general (i once hacked this):
[21:56] <hpk> start pypy
[21:56] <hpk> control-c
[21:56] <hpk> import py
[21:56] <hpk> print py.code.Source(space.div)
[21:56] <hpk> print py.code.Source(spave.floordiv)
[21:57] <hpk> but the two definitions don't show much a difference that would warrant your described behaviour
[21:57] <hpk> i would expect both to drag in the whole interpreter
[21:57] <stakkars> yes they would, but targetpypy1 uses this trick:
[21:58] <stakkars> exprargs, expr, miniglobals, fallback = (
[21:58] <stakkars> mm.install_not_sliced(space.model.typeorder, baked_perform_call=False))
[21:58] <stakkars> func = stdtypedef.make_perform_trampoline('__mm_'+name,
[21:58] <stakkars> exprargs, expr, miniglobals,
[21:58] <stakkars> mm)
[22:00] <stakkars> ah, thanks for your hint. Well, it's the generated binop wrapper code.
[22:00] <stakkars> we have so many indirections here that I'm quite desperate.
[22:02] <stakkars> HA!
[22:02] <stakkars> I even removed the floordiv implementation and it barfs.
[22:02] <stakkars> So I need to find some irregularity in object space setup.
[22:03] <stakkars> anyway, thanks.
[22:03] <stakkars> let's look into the win32 stuff tomorrow.
[22:04] <hpk> sure
[22:04] <hpk> i am heading off to other activities anyway
[22:04] <hpk> and indeed the descriptor-indirections are crazy stuff
[22:04] <hpk> but it's very similar to what cpython does :-)
[22:05] <hpk> class X(object):
[22:05] <stakkars> I just cannot figure out the difference of the setup here.
[22:05] <hpk> __class__ = list
[22:05] <hpk> isinstance(X(), list)
[22:05] <hpk> True
[22:06] <hpk> stakkars: yes, i guess it's easiest if you talk to Samuele who did it i think
[22:11] <stakkars> oh I get an idea!
[22:11] <stakkars> not sure yet, but these seemto be those which have no long implementation ???
[22:12] <stakkars> nope, they have :-(
[22:15] <hpk> ok, now i am really going, see you!
[22:16] <stakkars> buy
[22:16] hpk (~hpk@merlinux.de) left irc: "Client Exiting"
----- silence for 17 minutes ----- [22:33] pedronis (~Samuele_P@c-3a8b70d5.022-54-67626719.cust.bredbandsbolaget.se) joined #pypy.
[22:34] arigo (~arigo@c-3a8b70d5.022-54-67626719.cust.bredbandsbolaget.se) joined #pypy.
[22:46] <stakkars> hi! How was it? (whateverit was)
[22:47] <stakkars> pedronis: I've spent hours on multimethods. hpk pointed me to you that you might know it.
[22:50] <pedronis> I might know.
[22:51] <stakkars> I was trying lots of operators with a modified version of targetpypy1,
[22:51] <stakkars> that is, I take the multimethod and avoid generating the whole interpreter.
[22:51] <stakkars> the problem that I cannot find out about is:
[22:52] <stakkars> it works with a lot of operators, but not floordiv,truediv,lshift, rshift for instance.
[22:52] <stakkars> I looked into their implementations, but nothing special at all.
[22:52] <stakkars> so there must be *something* different somewhere. Do you have a clue?
[22:53] <stakkars> for instance, if I remove the buildcache2() thing, add mul sub div still work.
[22:53] <stakkars> with floordiv, I get a huge mess. So it touches something that I cannot find.
[22:53] <pedronis> well, one possibility if some of the implementation calls space.something
[22:54] <stakkars> I even removed the implementation, and it still happens. (replaced it by an empty function)
[22:54] <pedronis> I don't understand
[22:54] <pedronis> each multimethod is implemented in more than one place
[22:54] <pedronis> once per xxxobject.py file
[22:55] <stakkars> yes. ok, that's right. But there are not that many floordiv implementations at all.
[22:55] <stakkars> and the worst thing:
[22:55] <stakkars> the floordiv implementation is called by the div implementation.
[22:55] <stakkars> I can test div,but not floordiv.
[22:56] <arigo> stakkars: your question sounds a bit like "and if I put my leg behind my head in this way, what occurs?"
[22:56] <pedronis> stakkars:
[22:57] <stakkars> ??? I want to do my testing
[22:57] <pedronis> flaot floordive uses
[22:57] <arigo> it might point to a real problem, but that's a bit strange anyway
[22:57] <pedronis> space.getitem
[22:57] <pedronis> so that's one reason everything is brought in
[22:57] <stakkars> aahh, I understand
[22:57] <arigo> stakkars: it's better to write unit tests, probably
[22:58] <stakkars> arigo: hey, I must understand everything, this was simply a barrier :-)
[22:58] <stakkars> ok, so what happens is:
[22:58] <stakkars> not the way the things call each other, but the original method that was called triggers code in other modules
[22:59] <stakkars> and this doesspace stuff. phew many thanks.
[22:59] <arigo> I always have a file pypy/translator/x.py in my working copy
[22:59] <arigo> along the lines of "import pypy.translator.lotsofstuff"
[22:59] <arigo> "define a sample function"
[23:00] <arigo> "call the translator, and t.view()"
[23:00] <arigo> then I change the function or paste small examples in there
[23:00] <arigo> it's much easier to see what's going on than fishing around larger examples
[23:01] <stakkars> I do this all the time. My sample code is a single line. Just the wrong one :-)
[23:01] <stakkars> well, maybe I might have seen it in the flowgraph!
[23:02] <stakkars> but the problem was: using the wrong one liner, I don't get a flow graph before 20 minutes.
[23:02] <stakkars> thanks, anyway
[23:04] <stakkars> arigo: yes, I have done my unit tests. This was simply what I could not get right.
[23:06] <arigo> unit tests with no space at all, just checking the ovfcheck() function on some artificial examples
[23:06] <arigo> I fear that anything touching the targetpypy* is automatically disqualified as "unit test" and is a test of everything :-)
[23:08] <stakkars> as I said, I have the unit tests for that as addition to test_typed.py, that works just fine.
[23:09] <stakkars> And I thought to add the newly supported things to targetpypy, just to show them. Bad idea I guess .
[23:10] <stakkars> after all it was very helpful to me.
[23:10] <arigo> yes, I guess that's good
[23:10] <stakkars> I now have a much better understanding of multimethods, and how they grow a new caller" forme
[23:10] <arigo> starting to look around targetpypy usually gives a lot of unexpected insights :-)
[23:11] <stakkars> my big fear was that we have a deeper bug somewhere, that's a relief.
[23:11] <stakkars> There is a deepbug,but I have to check it on another platform.
[23:11] <stakkars> my first tests of rpystone worked fine, then they didn't.
[23:12] <stakkars> It turned out that they worked again after trashing somepyc files :-(
[23:12] <stakkars> the effect is that genc from time to time assignes a real number 5 to a structure, instead of a ginst_5
[23:13] <stakkars> but later on that and only if I can reproduce it.
[23:13] <arigo> ok
[23:14] <arigo> (strange indeed)
[23:14] <stakkars> even thought it was something with the pygame tools, but it isn't.
[23:14] <stakkars> still might be a stackless effect,so I'dbetter stoptalkin 8-)
[23:15] <stakkars> gonite
[23:15] <arigo> good night
[23:16] <stakkars> (btw. is there any objection if I fix the long impl?)
[23:16] <arigo> I guess not from me, I did one too
[23:16] <arigo> if you plan larger changes, have a word with Carl
[23:16] <arigo> just to avoid duplicate efforts
[23:16] <stakkars> well, basically removing the dependency fromlong
[23:16] <stakkars> ok, sure
[23:17] Action: stakkars schnarch
[23:30] cfbolz (~cfbolz@hdlb-d9b946f3.pool.mediaWays.net) joined #pypy.
[23:31] <cfbolz> stakkars: still there?
[23:34] <cfbolz> stakkars: (I hope you read the logs :-) go ahead with longs, I won't have time for that for a while.
[23:34] <cfbolz> bye
[23:35] cfbolz (~cfbolz@hdlb-d9b946f3.pool.mediaWays.net) left irc: "Verlassend"
[23:38] stakkars (~tismer@rosine165.inf.fu-berlin.de) left irc: Read error: 145 (Connection timed out)
[00:00] --- Thu May 5 2005