Geth MoarVM: briandfoy++ created pull request #759:
Update Unicode version		 06:42
brrt good * #moarvm 06:46
Geth MoarVM: deddb916cd | (brian d foy)++ (committed by Samantha McVey) | README.markdown
Update Unicode version cited in README

The moarvm.org page touts Unicode 10, so update the README to be consistent.		 07:08
MoarVM: 0406d43143 | (Samantha McVey)++ (committed using GitHub Web editor) | README.markdown
Merge pull request #759 from briandfoy/patch-1

Update Unicode version cited in README
samcv Never taken advantage of the fact i can edit peoples commits before merge. I think i'll do that more often to adjust wording in commit titles 07:09
dogbert2 o/ 10:14
dogbert2 wonders if jnthn has any brane today :)
jnthn also :) 10:38
nwc10 well, we have snow 10:39
no idea about braaaaaaaaaaaanes
jnthn Darn, the snow missed here. 10:40
dogbert2 the snow where I'm at melted away
ilmari it was down to freezing here overnight, but no snow, and it's up to 3°C now 10:48
jnthn works on a Cro middleware task as a warm-up :) 10:49
Well, that went to plan :) 11:13
dogbert2 an easy fix then ? 11:17
jnthn Was actually implementing a feature, for once, rather than fixing something ;) 11:20
dogbert2 cool 11:21
brrt good *, btw 11:29
dogbert2 hello brrt 11:30
brrt i note that i, for one, do not have any brane left
they were all out this morning
dogbert2 perhaps jnthn stole your brane :)
brrt i blame the lack of snow 11:31
jnthn Grr, this inlines/handlers thing is going to be quite tricky 11:32
travis-ci MoarVM build errored. Samantha McVey 'Merge pull request #759 from briandfoy/patch-1 11:43
travis-ci.org/MoarVM/MoarVM/builds/309367478 github.com/MoarVM/MoarVM/compare/d...06d431437b
jnthn Really so. Just discovering the number of places that assume a flat table. 11:46
And noting that each of them will get more complex. 11:49
Think I'll see if I can find an alternate solution that's going to be less invasive, now I've seen the cost of my original idea 11:52
jnthn lunch & 12:08
jnthn om nom borsch 13:16
So I think I have a far easier to implement alternative solution
nwc10 \o/
jnthn We poke entries into the handlers extent list that mark the boundaries of an inline
Thus the handler walker knows when it's crossing such a boundary 13:17
That provides a way forward on the "skip the caller, then look for outer" issue 13:18
If that extent list entry were to also contain an indication of which inline we just crossed the boundary of, then we can use that to do lookups to decide if this is or isn't a decent outer 13:19
src/strings/unicode.c: In function ‘MVM_unicode_get_property_str’: 13:24
src/strings/unicode.c:72236:83: warning: pointer/integer type mismatch in conditional expression ? Bidi_Mirroring_Glyph_enums[0] : Bidi_Mirroring_Glyph_enums[result_val]) : bogus;
samcv: ^
brrt jnthn, i'm not sure if i understood that correctly 13:48
jnthn Which bit? :) 14:03
brrt the entries-into-the-handlers-extent-list-that-mark-the-boundaries-of-an-inline
jnthn Oh 14:04
It's just putting extra rows into the handlers table
So that when we walk it, we can understand that we are leaving a region of code that was inlined
Geth MoarVM/inline-lastexpayload: c6bee06042 | (Jonathan Worthington)++ | 2 files
Remove :noinline on lastexpayload
 17:18
MoarVM/inline-lastexpayload: f60ceaa07c | (Jonathan Worthington)++ | 4 files
Eliminate inlined_and_not_lexical flag

This isn't sufficient, and wasn't being set in the right way anyway, thus hiding another bug. It will be replaced with a more correct and more precise mechanism.
MoarVM/inline-lastexpayload: db85b77a26 | (Jonathan Worthington)++ | 4 files
Re-work lexical handler lookup with inlines

The previous mechanism was a total cheat, and only survived because marking lastexpayload as :noinline meant we never ran into any of the cases that it didn't handle. This also prevented a lot of desirable inlinings from taking place. This corrects the handling of lexical handler resolution to work with inlines.		 17:28
jnthn That took some doing. 17:30
BTW, commit message is short but the code contains some details on things 17:34
Uh, code comments
m: say 5.27 / 6.49 17:38
camelia 0.812018
jnthn m: say 8.47 / 9.62 17:39
camelia 0.880457
jnthn Hm, 19% and 12% performance win on a couple of array-related benchmarks because postcircumfix:<[ ]> now inlines 17:40
ilmari nice 17:43
Geth MoarVM/inline-lastexpayload: 04e291170e | (Jonathan Worthington)++ | 2 files
Allow throwpayloadlex to be inlined

The limitations that prevented this before are now gone.		 17:45
jnthn One more win too
That one probably helps NQP code more than anything
Maybe compilation gets a tad faster
Spesh stressing seems happy 17:46
Geth MoarVM/master: 4 commits pushed by (Jonathan Worthington)++
jnthn So, there's the merge
jnthn wanders home 17:49
timotimo fantastic! 17:54
timotimo sees a ~27 line comment 17:55
Geth MoarVM: fac29990d1 | (Timo Paulssen)++ | src/profiler/heapsnapshot.c
revert accidental half-inclusion of incremental heap snapshot data
 17:56
travis-ci MoarVM build failed. Jonathan Worthington 'Allow throwpayloadlex to be inlined 21:45
travis-ci.org/MoarVM/MoarVM/builds/309632662 github.com/MoarVM/MoarVM/compare/0...e291170ec6
jnthn Tripped over the version bump 22:54
timotimo hm, we should be able to recognize that kind of failure with our bot, except the bot isn't in here 23:12
buggable, i think it is
wait, buggable *is* here?

Powered by Cro and the Raku® Programming Language.

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

Thank you!