Geth arVM: d41ccf81b9 | (Daniel Green)++ | src/jit/graph.c
JIT MVM_radix() like MVM_bigint_radix()

This means nqp::radix() will be JITted like nqp::radix_I() already is.		 10:10
arVM: a5749a80b8 | (Jonathan Worthington)++ | src/jit/graph.c
Merge pull request #514 from MasterDuke17/master

JIT MVM_radix() like MVM_bigint_radix()
arVM: d9e3969eea | (Samantha McVey)++ | tools/ucd2c.pl
Add shebang #!/usr/bin/env perl to ucd2c.pl
 10:13
arVM: fc3a77d27c | (Samantha McVey)++ | 4 files
Make radix 50% faster with Unicode Nd, add East_Asian_Width prop.

Don't use the general category which is slow and requires string operations. Check that Numeric_Type=Decimal
This should also speed up radix in general when it is given non A-Z, a-z or 0-9 codepoints
This also adds support for the East_Asian_Width property.
arVM: 0171f1731c | (Jonathan Worthington)++ | 4 files
Merge pull request #515 from samcv/radix_speed

Make radix 50% faster with Unicode Nd, add East_Asian_Width prop.
samcv kewl
jnthn Looked good to me :) 10:14
Geth arVM: 6043f9e42b | (Jonathan Worthington)++ | src/io/syncsocket.c
Fix memory leaks in synchronous sockets.
 13:13
arVM: 8b195f641c | (Jonathan Worthington)++ | src/io/syncsocket.c
Free socket handle in on_close callback.

This is the time it's ready to be freed and will be used no longer.
arVM: 17acfbe617 | (Jonathan Worthington)++ | src/io/syncsocket.c
Fix crashes in socket connect error handling.

Don't prematurely free, and don't longjmp over libuv internals.
Geth arVM: 9a179f275c | (Jonathan Worthington)++ | 2 files
Don't leak memory in STable repossession.

With this, `perl6-m -e ''` is free of leaks, which will make it easier to spot real leaks that need attention.		 14:14
arVM: 8f9de6fd06 | (Jonathan Worthington)++ | src/6model/serialization.c
Missing MVMROOT around a lock acquisition.
ilmari death to leeks! 15:35
nwc10 om nom nom 15:42
brrt .tell jnthn what also leaks (at least indirectly) is the memory for the extension ops in the JIT 16:25
yoleaux2 brrt: I'll pass your message to jnthn.
brrt .tell jnthn the solution for that (and a bunch of other things) is to have a data segment and stash the extension ops memory there, which is something i've been meaning to implement 16:26
yoleaux2 brrt: I'll pass your message to jnthn.
jnthn brrt: Yeah, we don't hit that in a straight ./perl6-m -e '' 16:28
But yes, I've seen the odd JIT leak show up and wondered if it might be those. Good to know it is. 16:29
brrt well, that's what i think it is, might be other things
better said: I know it's those
egh, better said: I know those leak, there may be other leaks as well 16:30
Geth MoarVM: 70d4bd53e6 | (Jonathan Worthington)++ | src/6model/reprs/NFA.c
Refuse to form an NFA with a zeroed to-state.

These will explode during evaluation due to reading memory out of bounds.		 18:44
MasterDuke samcv: did you see my comment on your PR? could/should the same change be made for MVM_bigint_radix? 23:26
samcv yeah i saw that
yeah and it can
MasterDuke cool, didn't want the thought to get lost 23:27

Powered by Cro and the Raku® Programming Language.

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

Thank you!