00:50 lue joined 02:45 jimmy2 joined 03:55 JimmyZ_ left 03:56 JimmyZ_ joined
dalek arVM/fixes: c667d60 | (Rob Hoelz)++ | src/io/syncfile.c:
Don't use filename to determine EOFness

  See github.com/MoarVM/MoarVM/issues/160		 05:21
arVM/fixes: 9af3581 | (Rob Hoelz)++ | src/strings/utf16.c:
Allow passing of length = -1 to MVM_string_utf16_encode_substr

  See github.com/MoarVM/MoarVM/issues/161
05:44 JimmyZ_ joined 07:40 rurban_ joined 07:43 FROGGS joined 08:10 zakharyas joined 08:57 brrt joined 09:32 woolfy left 09:33 woolfy joined 11:02 brrt left
carlin github.com/MoarVM/MoarVM/pull/163 11:12
FROGGS carlin: the patch to libtommath should be reported upstream also 11:30
and should be committed to nqp/3rdparty as well
carlin nqp pull req at github.com/perl6/nqp/pull/206 11:51
would libtommath take these patches?
dalek arVM: 27432fc | Carlin++ | / (2 files):
do not call srand() if not using rand()

there is no need to call srand on the platforms that arc4random is used this silences the warning OpenBSD's ld gives when it sees srand called		 11:56
arVM: 14aa33f | jonathan++ | / (2 files):
Merge pull request #163 from carbin/arc

do not call srand() if not using rand()
FROGGS carlin: eventually... they took PR's from us but these sleep in a development branch for two years now 11:58
carlin: but please send the PR anyway
carlin++ 12:00
carlin okay I might as well 12:04
12:08 brrt joined
brrt jnthn: could not perchance replicate my failing test cases? otherwise i'm going to write it off to OS limits of some sort 12:09
12:12 Util joined
hoelzro good moarning, #moarvm 13:09
brrt \o hoelzro 13:12
hoelzro o/ brrt 13:13
carlin libtommath merged my pull request already, yay 14:39
FROGGS wow 14:40
brrt carlin++ 14:46
hoelzro has anyone scrutinized my latest PR? github.com/MoarVM/MoarVM/issues/161 15:23
I was wondering if there was something unexpected I would be breaking
nwc10 I tested it, and I think it's an improvement on what we have
oh, wait, sorry, I looked at the other 15:24
18:33 FROGGS joined
jnthn hoelzro: I'm certainly good with 9af3581 18:43
hoelzro jnthn: cool, what are your thoughts on c667d60?
jnthn FROGGS: I think (sorry if wrong) that it might have been you who added the code being removed in github.com/MoarVM/MoarVM/commit/c6...43e2e95dee ? 18:45
hoelzro: My thoughts are "I'm sure I didn't put that code there, I want to understnad what led to it" :)
FROGGS the what?
jnthn Hang on, I'm doing git blame :) 18:46
FROGGS jnthn: I put the code in there that got removed in that diff, aye
jnthn Yeah, git blame thinks so too 18:47
FROGGS and it solved an issue back then... 18:48
:o)
jnthn Hmm
Can you remember what?
4a8dac66 seems relevant
FROGGS well, data->fd was uninitialized for some reason
jnthn a109e8d7 also 18:49
hoelzro jnthn: that's why I didn't just push it =)
if the check for data->filename were to remain, I would argue that it should be considered only if data->fd were uninit'd 18:50
FROGGS I think the code that caused an issue was: perl6 -e 'say lines' < foo.txt 18:51
or so
hoelzro builds and tries 18:53
nwc10 with the change, the EOF test fails on an unseekable file
jnthn FROGGS++ hoelzro++
nwc10 eg
./perl6-m -e '.say for lines' <(cat /etc/motd )
but the change is better than the status quo
(where that also fails, for some reason)
sadly, I believe that the only "reliable" way to do EOF is "try to read a byte" 18:54
hoelzro agreed
nwc10 which, obviously, has issues about blocking
hoelzro oh, is that why that stat == seek_pos check is done?
I read that last night, and it really bugged me
FROGGS hoelzro: that's the ticket that resulted in my patch: rt.perl.org/Ticket/Display.html?id=121720
hoelzro FROGGS: thanks! 18:55
nwc10 I don't have a great answer for any of this - I'm not familiar with the MoarVM code, nor with libuv
nor with what flexibility there is
nor with how to be portable to both POSIX and Win32
(if not further)
but "improvement" is better than "Status quo, awaiting perfection"
hoelzro I wouldn't mind having some regression tests for both of these behaviors; where do those belong? 18:56
nwc10 (I don't know) 18:58
jnthn In roast feels sensible to me, somewhere under S16 I guess 18:59
hoelzro FROGGS: good news! that behavior doesn't happen with my patch =D 19:00
FROGGS hoelzro: nice!
jnthn nwc10: About the "fails with unseekable file" - did it also fail before the patch? 19:04
nwc10 jnthn: yes, differently
so I think that the patch is an improvement, as I'm not aware of anything that it causes to regress
jnthn Right, i first understood what you wrote as meaning "it's a regression2 19:05
nwc10 ah sorry, no, wasn't clear
jnthn But yeah, if it's not, and it doesn't bust the RT causing the code we're removing, then I think I'm good with it
May be worth verifying on OSX, unless one of you is on that...
nwc10 and as well as the specific test case, the exisitng approach is a definate race condition, hence a security(ish) bug waiting to happen
nwc10 was not 19:06
jnthn Oh, moritz++ filed the bug, and I am sure he'd not be on OSX... :)
So I'm good with verifying the bug is still gone on Linux. :) 19:07
So, +1 from me to hoelzro++ patch too, unless anybody has any objections. 19:08
nwc10 to be very clear, I tried my little test with the <( ) shell syntax on linux with and without
I've not tried the specific test case that it attempted to fix
but from reading the code, I think that it's an improvement 19:09
19:09 FROGGS_ joined
hoelzro alright, I'll push those then 19:10
I should probably test different encodings too
dalek arVM: 3851372 | (Rob Hoelz)++ | src/io/syncfile.c:
Don't use filename to determine EOFness

  See github.com/MoarVM/MoarVM/issues/160		 19:11
arVM: babd3dc | (Rob Hoelz)++ | src/strings/utf16.c:
Allow passing of length = -1 to MVM_string_utf16_encode_substr

  See github.com/MoarVM/MoarVM/issues/161
hoelzro I think we could also ditch MVM_file_stat_follow_symlink as well 19:12
actually, I think that function will spin forever on a self referential symlink
jnthn You can have...self-referential symlinks? :P 19:17
FROGGS_ hehe 19:18
some of us perhaps, yeah :D
hoelzro ln -s bar bar
jnthn but...but why? :P 19:19
hoelzro no reason, but it's possible =)
nwc10 "it's a trap" 19:20
hoelzro you could also have foo -> bar -> baz -> foo 19:22
japhb jnthn: I suspect that since *nix can't cheaply solve the problem of arbitrary symlink loops, it simply throws up its hands and refuses to give a 90% solution that might leave you with a false sense of security. (But I don't actually know the history.)
What hoelzro said
hoelzro what confuses me about MVM_file_stat_follow_symlink is that stat follow sym links 19:23
so it seems redundant 19:24
but that's my Unix bias showing
FROGGS_ as I've learned there is stat and then lstat
and the one we used did not follow symlinks
hoelzro lstat doesn't 19:25
jnthn japhb: I have a windows bias, so I'm just like, "bah, you crazy kids with your symlinks" :)
FROGGS_ that's why "foo".IO.s reported 3, since the name has three bytes
jnthn: well, you can mount volumes in a folder nowadays... perhaps you get symlinks by 2020 19:26
jnthn :P
nwc10 Apple re-implemented their OS onto FreeBSD (on a Mach micro-kernel) 19:29
So maybe MS should re-implement theirs onto OpenBSD
and stun the security naysayers 19:30
japhb
.oO( Unix: Three decades ahead of Windows in basic functionality, three decades behind in video driver stability)
tadzik hehehe *sniff*
FROGGS_ I don't need a video driver to ssh into another box :o) 19:35
jnthn Funny thing is, when the video driver crashes on Windows these days - which is for me about 2-3 times a year - it's just like "oh, your video driver crashed, so I respawned it for you, kthxbai" :) 19:47
tadzik heh, I see games crashing on linux every now and then 19:52
and dmesg just says "there was a segfault, it's gone now"
jnthn brrt: No, I can't get the NQP tests to blow up on Windows... 20:30
21:46 dalek2 joined 21:47 d4l3k2_ joined
timotimo holy fuck github has source code highlighting in diffs now 22:15
<3
japhb The intro to your statement made me think of The Illuminatus! Trilogy. 22:19
timotimo i've only read the first book - and in german 22:24
i had a hard time understanding it
but i think it's supposed to not be understood
japhb I would easily believe that a translation would make it even worse. 22:25
timotimo hmm 22:26
japhb I read it after I saw a review that said essentially "after 100 pages, I wondered why anyone would read this, after 200 I thought it had potential, after 300 I couldn't put it down". Which was essentially my experience. :-)
timotimo hah! :D 22:28
[Coke] If you're looking for other mind-rending books to read, I just finished Kraken by China Miéville. 22:29
timotimo i may give illuminatus! another try 22:32
tadzik has just finished VOY 22:58
japhb VOY? 23:00
tadzik Voyager 23:01
japhb The book or the TV show? 23:02
tadzik oh, there's a book? :) No, just the show 23:06
japhb So what did you think of the ending? :-) 23:13
23:13 colomon_ joined
tadzik it is a bit of a deus ex machina 23:15
bit aside from that, I liked it :)
it surprised me a few times
[Coke] Ugh, I watched every ep on dvd from netflix over the course of a few months. 23:26
japhb [Coke]: No streaming? 23:29
I've got to say I'm impressed with modern streaming encodings. Even with our crappy internet connection, streaming video usually looks way better than DVD video. 23:30
japhb wonders if Netflix ever reversed their decision way back when to stop carrying Blu-Ray disks 23:31
[Coke] japhb: this was about 6 years ago. 23:46
wasn't available for streaming then.
japhb Ah, gotcha. 23:49
[Coke] plus with dvds you can watch them at work on your break! 23:53

Powered by Cro and the Raku® Programming Language.

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

Thank you!