% mdiep has left mdiep!mdiep@bursley-216-65.reshall.umich.edu
% cognominal has left cognominal!~stef@cac94-1-82-67-232-89.fbx.proxad.net
% cognominal has joined #parrotsketch
% mdiep has joined #parrotsketch
% particle has joined #parrotsketch
% leo has joined #parrotsketch
% allison has joined #parrotsketch
% Nicholas has joined #parrotsketch
<Nicholas> Hi
<obra> hello, all
<allison> hi
<mdiep> hello
<obra> leo, chip, pmichaud: you here?
<leo> hi all
* leo was just dropping noodles into hot water
<obra> we'll give the others another two minutes or so, then get cracking.
<particle> hello
<obra> hey Jerry.
<obra> ok then. let's get going
<obra> Leo: what's new and exciting?
<leo> - lot of code cleanup, bug fixes, and improvements
<leo> - e.g. fixed the memory corruption bug that stopped tcl
<leo> - shootout improvements and a bit of optimization:
<leo>   http://shootout.alioth.debian.org/sandbox/parrot.php
<leo> - an the bad news is: Strings are heavily broken due to COW copying too much
<leo>   (some shootout benchmarks just break or are almost unimplemetable)
<leo> that's it
<obra> Ok. whatcha blocking on?
<leo> Strings need some redesign
<leo> I need chip for further decisions
<obra> Which decisions?
<obra> ;)
<leo> I've written some design pods and need an 'ok'
<obra> Are these full PDDs?
<leo> not full - more dev guidelines
<leo> they are on feather too
<leo> ~lt/dev-doc/
<obra> cool. also in svn?
<leo> not yet - I'd like to talk with chip first
<obra> Ok.
<obra> anything else?
<obra> if not, Allison: what's new?
<leo> in the meantime I'm continuing code cleanup and improvements e.g. Parrote_allocate* code
* allison just finished making oatmeal
<allison> I finished off the refactor to make Punie use a precompiled PGE grammar
<allison> Thanks to Patrick for answering some questions on that one.
<allison> Next step is another expansion, more operators, more argument types.
<allison> EOR
<obra> Anything impeding your fun?
<allison> Aside from not being able to eat anything more solid than oatmeal? No. :)
<obra> :/
<obra> ok. Jerry. what's up?
<particle> dir reorg is done. there's still some file/renames to happen, but those will just be little changes here and there.
<particle> after the holidays, i'll have more time to devote to the test suite, including a test plan.
<particle> in the meantime, i'll babysit windows and try to keep it compiling, and warning free
<obra> excellent
<particle> punie's test suite is taking up too much memory on windows, specifically t/past_*.t
<particle> i'll hunt down that as i can... perhaps it's a COW problem.
<obra> punie's?
<particle> yes.
<particle> punie, on my win32 box, takes up to 850MB to run the demo.
<leo> brr
<obra> ooh. allison: and _This_ is what you get for me harassing you to check your code into parrot svn. more bug reports ;)
<allison> particle: is it just the past tests, or is it also post tests?
<allison> yay! I like bug reports! :)
<particle> i'll check post tests now...
<obra> though "too much memory" implies it's actually running through. is it?
<particle> post, too. it does run, yes, because i have 1.5GB handy. but it fails on my 512MB machine.
<obra> that's still really cool. allison++ # punie lives
<allison> It does create some complex data structures internally, but they shouldn't be that large.
<particle> definitely unreasonably large.
<leo> could be really that mentione COW problem
<allison> Any suggestions on profiling PIR code?
<leo> which kind of profile?
<allison> I want memory usage, for one thing.
<leo> not much there except ./parrot -D1 foo.pir
<allison> But also benchmarks on the speed of various subroutines.
<allison> leo: Okay, I'll try it out.
<leo> only opcode granularity currently with -p (and then you guess where ;-)
<leo> but we definitely need better debug/profile code
% chromatic has joined #parrotsketch
<allison> Anyone feeling inspired to write Devel::DProf this week?
<allison> maybe later :)
<chromatic> Don't we need runloop-level hooks for that?
<allison> particle: I'll try it on some of my older machines and see if I can duplicate the problem. Let me know if fixing the COW problem fixes it.
<particle> allison: thanks, and will do.
<allison> particle: ooh, do you get the problem when running the language tests?
<allison> particularly io_print.t?
<particle> checking... prove t/io_print.t
<particle> no. that's less than 10MB
<allison> okay, that is weird
<allison> that's the one that should be a memory hog if anything is
<allison> the past_*.t and post_*.t tests only create a few sample objects.
<allison> io_print.t runs through a whole parse and compile
<allison> so it's the one that creates huge complex data structures
<allison> And quite a few of them.
<allison> weirdness, I'll look into it
<particle> perhaps i can create a small(er) test case based on one of the existing tests
<particle> i'll let you know if i can, but time is short this week.
<allison> ok, sounds good
<obra> ok. 
<obra> what's next?
<obra> or rather, who's next?
<obra> mdiep: you around?
<mdiep> yup
<mdiep> things picked up in tcl-land this week, with some things just getting committed during this meeting:
<mdiep> - Changed Getopt::Long so that it doesn't use PGE
<mdiep> - Change ParTcl to only load PGE if we actually need it (which speeds things up and gives smaller traces)
<leo> mdiep++
<mdiep> - Leo found the macro error based off an example I gave him. But we're still seeing memory corruption.
<mdiep> - We run our tests with -D40 now by default
<mdiep> - We fixed bugs in [for], [lappend], [lrange], [lsort], [unset]
<mdiep> - and added support for $argv and [string is]
<mdiep> - plus general code cleanup
<mdiep> We should be passing more of Tcl's test suite now, but I haven't run it again yet.
<obra> very cool
<obra> Has chip gotten back to you yet?
<mdiep> We really need the memory corruption to get sorted out. It really impedes development. And we need macros sorted out.
<mdiep> nope. I haven't heard from chip.
<leo> mdiep: which further mem corruption? 
<leo> where is the bug report ;-)
<mdiep> leo: if you run partcl's make test you should see it. There are ~4 tests that fail with IMCC parse errors as a result of the corruption.
<mdiep> I just checked in the code that exposed this. :-)
<leo> nope - you get a distinct error message
<leo> ah - wait
<mdiep> Oh, and we've got a "attempt to access code outside of current code segment" error in t/cmd_inline.t
<leo> namespace pollution i.e. using the same sub name in different namespace (I presume)
<leo> 4 errors due to macros, one due to that
<leo> nothing else is failing and you get an error message from imcc for all 5
<mdiep> using the same sub name in different namespaces is an error? that's a parrot bug, I presume.
<leo> consider it as an parrot bug, yes
<leo> there isn't a fast fix though
<mdiep> leo: only one of the [break] and one of the [continue] bugs give me "too many macros" now. The other two are IMCC parse errors.
<particle> mdiep: tcl build is failing on win. it's a pbc_merge issue, but i don't have the windows debug-fu yet to track it down. ...i should have thought of this earlier: jonathan worthington a) has windows, b) wrote pbc_merge. i'll ask him if he can look at it.
<leo> I have 4 times: Too many macros
<mdiep> anyway, we're blocking on these issues technically and they make development not fun. :-)
<leo> no one did answer to my mail on p6i WRT macros (no fun)
<mdiep> leo: I suspect you haven't rebuilt Getopt::Long and Tcl along with it
<leo> I just did 'make clean test' in tcl - yes
<obra> ok. leo: do you have a decent sense of what you need to make matt's issues go away?
<obra> (aside from a working build ;)
<mdiep> I was waiting for chip to answer leo's mail. I don't really care which way they're implemented, as long as they work. (there are a couple options)
<leo> well, and I don't change anything until I have some opinions ;-)
<obra> ok.
<mdiep> leo: remake runtime/parrot/library/Getopt/Long.pir to get the different errors in Tcl's make test
<obra> Moving along then.
<obra> As I suspect you've all seen by now, Nick has moved on from Fotango. He was working on Ponie as part of his job for Fotango. So now we're hunting for a new Ponie Pumpking.  If y'all know any good suckers, be sure to send them our way.
<obra> Nick: how are you enjoying your retirement?
<Nicholas> It's quiet.
<Nicholas> I'm missing the people.
<Nicholas> The call for pumpking featured in the list summary today, so that should extend its reach to a wider readership.
<Nicholas> Also, I did some stuff.
<Nicholas> I figured out how to do the iteration/reference collection/garbage collection.
<Nicholas> Rather than try to work out how to butcher the perl hash implementation to try to have the iteration behaviour that I needed, I realised that it's only needed as an interim until the stack walking is enabled after all the reference counting inconsistencies are found and fixed.
<Nicholas> And that I was only using 1 pointer in the ptr_table's struct. And that the hacked about ptr_table already iteration does exactly what it needs. So I could use the other slot to hold a reference count. *All* that was needed was to add a hook in to do the marking. So I committed a minimal PMC implementation with a mark VTABLE entry.
<Nicholas> All is happy, and the "real" (sane) reference counting PMC will be useful once sane GC is in place and only XS modules are really doing reference counting.
<Nicholas> (and aren't trying to delete PMCs during global destruction)
<Nicholas> gah. I bet that's not quite true. :-(
<obra> So, this "holiday" thing is really working out for you, then? ;)
<Nicholas> er. sort of. :-)
<obra> Glad to see that it's productive at least ;)
<Nicholas> I think I've figured out how to drop the memory usage of XS module constants and actually get them inlined. But that's not on topic here
<Nicholas> I'm sure my parents' dog feels that I'm not productive. I mean, I didn't even walk him once today. Or yesterday.
<obra> I presume you're blocking on "a replacement"?
<Nicholas> I could argue yes. Or, in theory, being employed to do it.
<Nicholas> (in some form that pays the rent, long term)
<obra> fair enough
<obra> chromatic: anything new and exciting?
<Nicholas> (particularly having just discovered that UK pension changes on A-day (next year) abolish carry back so I need to deal with this year's pension this tax year, which means spending more money sooner than I previously thought)
<chromatic> I checked in the Test::Builder and Test::More tests.
<chromatic> It's a horrible very bad hack, mostly because I only put #!../parrot at the start of both PIR files.
<chromatic> it works though.
<particle> even on windows, surprisingly.
<chromatic> It's because all of the test files run through Perl, which itself interprets the shebang line.
<chromatic> I don't think this is a perfect solution, but it let me get this done without rewiring t/harness.
<chromatic> So it was quick and it works.
<chromatic> I also fixed almost all of the compilation warnings I saw on Linux PPC.  One was in the JIT.
<chromatic> There's one remaining -- assigning a constant string to a non-constant struct member.  I'm not sure how worthwhile the fix is (in pbc_merge.c) because the easiest fix is a strdup().
<chromatic> But everything still passes here after all of that, so I'm happy.
<chromatic> EOT
<obra> neat!
<obra> blocking on anything?
<chromatic> I'm still playing with ideas for Parrot::Test -- just haven't really set them down in code yet.
<chromatic> Time, money, love.
<obra> *nod*
<chromatic> Pick three.
* obra sends lawyers drugs and guns
<obra> Right then. Other questions?
<particle> next meeting?
<leo> just a note to mdiep - all the same after rebuilding
<obra> oh. right. Next week seems like an awfully good time to take a week off.
<obra> So. let's reconvene on 2 Jan
<Nicholas> Neither a question nor on topic, but in the food games, I'm about to make beef burgers. Mmm, raw mince. :-)
<leo> mahlzeit nick ;-)
<obra> Enjoy, Nick.
<Nicholas> Get well soon, Allison.
<allison> thanks :)
<particle> happy new year, all.
<obra> Have a great couple of weeks, everybody. Keep up the hacking. Don't spend too much time with your families, neglecting the parrot ;) (I'm kidding. Really ;)
* particle feeds the monkey^Wparrot
<leo> nice holidays to all and a happy new year (if a new one starts in your calendar)
<obra> *wave*
<leo> see ya all
<allison> see ya
<Nicholas> au revoir
% chromatic has left #parrotsketch
% particle has left #parrotsketch
% Nicholas has left #parrotsketch
% leo has left #parrotsketch
% allison has left allison!~allison@sub17-30.member.dsl-only.net
% chip has left chip!~chip@feather.perl6.nl
% chip has joined #parrotsketch