Post closed tickets in your report. | Note: This channel is for our weekly status meetings (Tuesdays at 19:30 UTC); you probably want #parrot instead. | irclog: irclog.perlgeek.de/
Set by moderator on 12 October 2014.
07:09 rurban joined 07:10 rurban left 07:19 lizmat joined 08:21 lizmat joined 13:21 rurban joined 15:22 rurban1 joined 16:42 Chirag joined 17:03 rurban joined
Chirag #Pre-Report 18:58
*Started figuring out the autoboxes. Trying to use the trace flags. Currently, I am finding the implementation a little difficult to understand but will understand better for sure
.end
<PS - I was just wondering if the day for #ps can be changed? Maybe Friday same time? It actually falls in the middle of the week. But only if convenient and possible, else I am fine :)>
Util Pre-report: 19:28
* Smolder:
+ Worked with OSUOSL to reboot the Smolder box, which did not resolve the problem.
+ Found that the problem is cause by bad quotes in meta.yml 19:29
+ SSH'ed into Smolder, and debugged from that side.
- Now tracing why the quotes only occur on Linux system.
* Changing the #parrotsketch schedule
+ I have no objection to changing #ps to Friday at the same time.
+ However, I am surprised that Chirag does not ask for an earlier time.
- Isn't 19:30 UTC around midnight there?
.end
rurban Done: (again a lot)
* Merged several branches and fixed minor issues, mostly for perl6.
Optimize utf8_substr, sprintf counts bytes, fix sprintf width for unicode and
NaN/Inf/-Inf, smoke-me/newlines3-gh1087
Work:
* Prepared release
* smoke-me/null-call-gh1024 SEGV in t/compilers/imcc/syn/const.t test 41 (empty imm constant sub)
* rurban/coros-gh564: added auto-reset, not ready. Not sure if wanted.
We could initialize a coro to use autoreset semantics. (add new attrib or flag)
* rurban/coro-reset-gh564: added Coroutine.reset - ready
* build changes: use gdb, installablelibparrot, ... ready, but need to wait for my test machines
* rurban/lexqnames-gh1095 .lex quoting the simple way (another hack on top of more hacks - not ready)
rurban/lexqnames-gh1095-dev rewrote imcc string handling, proper quoting handling (not ready)
To be reviewed:
* smoke-me/darwin-gdb-gh1096
* rurban/darwin-build-gh1096
.end
Util Hello! Who all is in attendance? 19:30
rurban me
Util Hi, rurban
rurban Hi, we should do friday, but earlier. for me it's also very late
Util Earlier is fine with me, either Tuesday or Friday. 19:31
rurban Chirag wanted Friday, I'm ok with Friday if earlier. Friday is movie day for me very often 19:32
Util How much earlier? Range 0..6 hours earlier is fine here. 19:33
rurban the earlier the better
even more so for Chirag
Fr morning for you sounds good
moderator Post closed tickets in your report. | Note: This channel is for our weekly status meetings (Fridays at 13:30 UTC); you probably want #parrot instead. | irclog: irclog.perlgeek.de/ 19:35
rurban Like this? 19:35
that's 8:30am Central Time
Util Yes; -6h from current #ps time is 8:30AM here, but that is 6:30AM in California, so we might have to change it again during the next GSOC. 19:36
rurban and 5:30pm indian time
yes
Util Let's try it for a week, and see if anyone objects. I am not sure what time it will be for Coke and lizmat, for example.
rurban lizmat is my time 19:37
Util OK
rurban Coke no idea
I guess Eastern Time
Util Coke: Whenever you see this, please weigh in.
rurban I am just slowing down method calls a bit, a flag check for empty invocants to prevent SEGV's smoke-me/null-call-gh1024 19:38
I guess it's justified. I cannot detect this in imcc, not even in the post fixup. Maybe I'll find out later
Util One key advantage to Tuesdays was that it is always release day. We will need to monitor the channels more closely on release day now. 19:39
rurban Saturday is bug day. So one day before bug day is also good
I'll send an email about my coro reset work and questions. 19:40
Util A safe check now sounds wise, (gh1024)
rurban moar doesnot have coros, they said
so it's not urgent
Util How is Rakudo planning to handle Perl-6-level coros on Moar without support? 19:41
(just curious)
rurban No idea
Util OK
rurban Sounds perfect for lazy lists to me
but currently they have to catch an exception when the list is empty 19:42
Util Your email will be helpful; wider audience, and better medium for potential conversation.
rurban in the pir book the semantiocs is to autoreset, which does not work for lazy lists
normal behavior is to return nil, and signal the end somehow 19:43
I have 2 proposal's which are fine I think. It's just not ready for the next release
the other big problem, the biggest problem is unicode slowdown. github.com/parrot/parrot/issues/1097 19:44
Just fixed substr, but there must be more. My initial guess was imcc, but I don't see it in the profiler 19:45
I'll try more profiling with this
Util Unicode is manditory, but ICU is just a long-standing convenience.
rurban ICU is not the problem. the iterator was. and now I think it's either perl6 or imcc 19:46
Util just read the issue. nm about ICU
rurban or really parrot run-time. then it will be hard to improve
utf8_substr is still 10x higher than the next op 19:47
so this looks like the whole parser needs iterate over utf8 natively 19:49
which is not doable for us
I'll also profile moar for this case
Util Sounds good. I suspect that a deep dive into the iterator may yield other fruits; i.e. tickets/tasks for next year's GSoC 19:51
rurban Improving imcc is definitely one of my goals. -O2 was not that good with perl6 code. 19:52
besides the calling convention, and ext_call 19:53
and my build changes might bring in -lparrot for so's again. but only for —debugging 19:55
so that you don't have to recompile all so's when installing 19:56
just one, libparrot.dylib
linux/bsd is okay, just windows/cygwin/darwin is broken
*broken* : hard to debug
Util I am suprised that libparrot.dylib does not work with gdb. (Do I understand/remember correctly that that is the driver for the build change?) 19:57
rurban yes, gdb mostly
I cannot work on branches alone, I will always have to apply my temp fixes 19:58
And I also want —debugging=-g3 to work, and —ccld=gcc-mp-4.9
Util I do not object to static compilation being restored as an option, as long as it stays *optional* 19:59
rurban I found out that a static parrot_old behaves different to a dynamic parrot in the order of initialization 20:00
immediate sub fixups are different. very strange
but for debugging parrot_old is much easier, faster to rebuild 20:01
I also tuned the Makefile a bit
make all should build all, not just the core.
and maybe check a Configure flag to turn of sharedlib 20:02
even --parrot_is_shared has no negation
Util This is all in smoke-me/darwin-gdb-gh1096 and rurban/darwin-build-gh1096 ?
rurban yes
the simply stuff in smoke-me/darwin-gdb-gh1096 20:03
Util OK
rurban much more in rurban/darwin-build-gh1096
target is 7.0.0
Util OK
rurban support --parrot_is_shared=0 should be doable also
or =no 20:04
and fix rpath on darwin. need to wait for my ppc laptop for this
so for the build changes I wait for my machines to actually test everything
coros could go in earlier 20:05
imcc unicode and quoting issues also earlier
and yes, one more issue I'll bring up in an email: Illegal escape sequences 20:06
do not silently skip illegal escape sequences anymore, die
Util Absolutely!
rurban This is esp for imcc bugs, when \\o is transformed to o
Util rurban++
rurban FROGGS came with this last week 20:07
Util Right
rurban this is in the lexqnames branches
I'm not sure how to deprecate the silently skip \\ behavior 20:08
I'll ask perl6. They are fighting with this, and they need to change their quoting style to single quotes for .lex now
Util I don't think that it needs deprecation. Was the "silently skip" behavior documented? 20:09
rurban single-quotes are very consistent within imcc, but not every statement accepts single quotes
I hope this will be ready until saturday
how about —debugging off by default? 20:10
github.com/parrot/parrot/issues/1099
I think this go in into 6.9.0 20:12
Util Why make that change?
rurban it's odd to have —debugging and not using it does the same
and I want to expand macros via -g3 20:13
gdb> macro expand Obj_is_PMC_TEST(object)
Util Are you saying that there are currently *no* differences in Configure'ing with vs without --debugging? 20:14
rurban I don't see any
—debugging is on by default
which is fine for a <1.0 release 20:15
but not if >1.0
init::optimize might be a bit different I guess
Util Let me ask that differently:
Are you saying that there are currently *no* differences in Configure'ing with --debugging=1 vs --debugging=0 ? 20:16
rurban lemme check...
yes, there's none 20:17
—debugging=0 does something 20:19
but leaving out —debugging is the same as with —debugging
—debugging=0 unsets cc_debug
Util That is my point. 20:20
rurban it's convenient for devs, maybe that's why
Util If you change the default for plain run of `perl Configure.pl`, that is OK, as long as: 20:21
20:21 kjs joined
Util A. it is documented, and 20:21
rurban —optimize does something like —debugging=0 plus -DNDEBUG
Util B. our clients (Rakudo) can change *their* invocation of `perl Configure.pl ...` to recover the debug behaviour, if they choose. 20:22
rurban so the default is to set -g and leave out -DNDEBUG (use assertions)
perl6 is fine. they just use —optimize
Util So, if you co-ordinate with the #perl6 team, then I have no further objection. 20:23
rurban my new default would be to set debugging to 0, but leave -g and leave out -DNDEBUG
Util They may be fine, but please make an explicit communication to them about this.
rurban so the same build flags, just change the misleading config hash entry
ok
Util Thanks! 20:24
Anything else to discuss? From anyone?
rurban late already
nope
Util Meeting adjourned.
20:24 rurban left
Coke When the meetings are has no impact on me at this point. No worries. 20:29
Please update the google calendar event, though. 20:30

Powered by Cro and the Raku® Programming Language.

Please report any issues / comments / feature requests as an issue on App::Raku::Log.

Thank you!