Parrot 6.6.0 "Parrothead" | parrot.org/ | Log: irclog.perlgeek.de/parrot | #parrotsketch meeting Tuesday 19:30 UTC
Set by moderator on 16 July 2014.
01:16 FROGGS_ joined 02:21 kid51 joined 04:09 estrabd joined 04:15 perlfan joined 04:51 bighugedog joined 06:36 basiliscos joined 07:31 kjs joined 08:08 basiliscos joined 08:32 wagle joined 09:19 kjs joined 09:30 Psyche^ joined 11:05 kid51 joined 13:29 rurban1 joined 13:31 bluescreen joined 14:14 khisanth_ joined 15:02 Chirag joined 17:31 rurban1 joined
Chirag rurban: gist.github.com/ZYROz/2b709dc74bd354143f81 18:10
But my log.data looks very unreliable
18:32 rurban1 joined
rurban 152 vs 67 secs :) 18:41
rather 151617.489213,task-clock vs 152306.055310,task-clock 18:42
so is RELEASE_6_6_0-0-gad40760 static call_obj and RELEASE_6_6_0-1-g71d5afe the direct call? can we have a diff between those two commits? 18:43
Util #ps time 19:32
Chirag no.. I am still working on the direct calls part.. This is after changing pmc_callcontext.h only 19:33
19:33 rurban1 joined
Chirag diff does not work since the .h file is not monitored, I guess .. But basically all the macros are changed... 19:41
Util To get a diff of an untracked file, build the software in its unmodified form, then copy the untracked file to a new name, then modify, rebuild, and run "diff -u newName.h currentName.h". 19:43
Chirag oh.. thanks for that! 19:44
rurban I see 20:10
Maybe it's better to add this hack and push it. so it's easier to repro 20:11
Chirag hmm..
rurban we just need to remember to remove it later 20:12
Chirag ok..
rurban and maybe make it readonly so pmc2c will not overwrite it 20:13
chmod -w
Chirag pushed 20:18
20:18 davidfetter joined 20:35 rurban1 joined
rurban I also prepared a faster callcontext now, just not expanded the vtables calls yet, only ATTR 20:37
all tests pass
dalek rrot/rurban/pcc-gh1083: 71d5afe | ZYROz++ | src/call/pcc.c:
[pcc] inline pcc_add_invocant and its attr accessor
 20:39
rrot/rurban/pcc-gh1083: 56e2d00 | rurban++ | / (2 files):
[pcc] temporarily add two hand-optimized callcontext pmc files

see gh #1083
rrot/rurban/pcc-gh1083: 7b7a019 | ZYROz++ | src/call/pcc.c:
[pcc] inline pcc_add_invocant and its attr accessor
 20:40
rrot/rurban/pcc-gh1083: 4353c37 | rurban++ | / (2 files):
[pcc] temporarily add two hand-optimized callcontext pmc files

see gh #1083
rurban my load is still >5 20:50
( +- 9.17% ) 20:51
3% faster 21:10
dalek rrot/rurban/pcc-gh1083: 7faa17b | rurban++ | / (2 files):
[pcc] more hand-edited changes GH #1083
rrot/rurban/pcc-gh1083: 970c54b | rurban++ | src/call/ (2 files):
[pcc] actually use now the direct methods GH #1083

and do not use the CallContext vtable methods anymore. also inline one Continuation_set_pointer call.
all tests still passing, 2.98% faster.
rurban so lets change pmc2c to use the faster CallContext calls
Chirag: can you try also my changes? 21:11
21:13 basiliscos joined 21:35 rurban1 joined
rurban FROGGS_: I'll rather fix tools/dev/gen_charset_tables.pl also. just the u_hasBinaryProperty(codepoint, UCHAR_LINE_BREAK) looks good then 21:44
FROGGS_ yeah, probably makes sense
rurban - $ret |= 0x0020 if $chr =~ /^[[:space:]\\x85\\xa0]$/; # CCLASS_WHITESPACE
+ $ret |= 0x0020 if $chr =~ /^[[:space:]\\x85\\xa0\\x2028\\x2029]$/;# CCLASS_WHITESPACE
- $ret |= 0x1000 if $chr =~ /^[\\n\\r\\f\\x85]$/; # CCLASS_NEWLINE 21:45
+ $ret |= 0x1000 if $chr =~ /^[\\n\\r\\f\\x85\\x2028\\x2029]$/;# CCLASS_NEWLINE
I'll check
FROGGS_ rurban: \\x2028 does not make much sense there, because this is only used to generate the properties of the first 256 codepoints
also, \\x2028 will match " 28" 21:46
rurban oops
FROGGS_ and as I said, >\\xFF is a noop there
for some reason my libicu does that think that \\x2028 and \\x2029 are newlines 21:47
that's why I hardcoded these two
rurban so we need m//u and properly escape them
oh, I see 0..255 only 21:48
FROGGS_ well, only if that script would create a table for all codepoints :o)
so you can leave out the \\x2028 and \\x2029 there
rurban "Create a whole character table" is a bit overly enhusiastic then 21:49
Uh, and Parrot_ascii_typetable is missing, and the generated table deviates wildy 21:53
ascii was changed to fixed8 with 505fdc4c9a48fe90d17 21:54
ok, /^[[:space:]\\x85\\xa0]$/ is needed, already included. now looking for a POSIX newline class... 22:04
is *not* needed 22:08
FROGGS_: 11 is a newline (VT)? I think not, POSIX thinks neither. only 12 and 13 are 22:11
oh, with Word it served as a newline 22:12
(powerpoint slides are using it)
LINE TABULATION 22:13
FROGGS_ # Line_Break=Mandatory_Break
000B..000C ; BK # Cc [2] <control-000B>..<control-000C>
rurban so it's $ret |= 0x1000 if $chr =~ /^[\\v\\n\\r\\f\\x85]$/; # CCLASS_NEWLINE 22:15
then
right?
So I've got discrepancies with 215 and 247 in [[:punct:]] 22:17
FROGGS_ /^[\\v\\n\\r\\f\\x85]$/ is correct
rurban 160,168,176,184,190 also
and some more 22:18
FROGGS_ :/
rurban I think I'll rather trust libc for this
FROGGS_ gnight 22:21
rurban thanks again
FROGGS_ ohh, I have to thank you :o)
22:36 rurban1 joined
dalek rrot/FROGGS/newlines2-gh1087: 7789e9f | (Tobias Leich)++ | src/string/encoding/ (2 files):
RT #122341 treat all of 0x0a-0x0d,\\x85,\\x2028,\\x2029 as newlines

The script tools/dev/gen_charset_tables.pl was not used because it removes character properties of chars in the range of 0x81 to 0xFF. Additionally, u_iscclass now checks for characters with enum_cclass_newline property, which it did not do at all before.
Signed-off-by: Reini Urban <rurban@cpanel.net>		 22:50
rrot/FROGGS/newlines2-gh1087: c1d57d2 | rurban++ | / (4 files):
regenerate src/string/encoding/tables.c

so that tools/dev/gen_charset_tables.pl is kept up-to-date.
  (it was not)
Added \\v to CCLASS_NEWLINE manually (confirmed).
  \\x85\\xa0 already is in the whitespace cclass.
Removed the creation of the now defunct ascii table.
Beware: My new up-to-date libc removed the [[:punct::]] class of several chars > 160.
22:57 rurban1 joined 23:50 kid51 joined

Powered by Cro and the Raku® Programming Language.

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

Thank you!