00:37 dalek joined 00:43 virtualsue joined
orbus greetings moarvm folks 01:46
was curious if anyone knows what the current status of moar on sparc is
I spent some time last week getting it to build on solaris 10, which I finally got it to
but it dumps core when I try to build nqp
I had to build a newer gcc to build moar, and I'm thinking the gcc build may be no good - but figured I'd come ask if this *should* work, or if it's a no-go regardless
01:48 ilbot3 joined
TEttinger orbus, I think it's a little sleepy in here right now. that sounds like an impressive amount of work you've put in though, I imagine there's a way to make it work 02:46
orbus no worries - I can ask later 02:47
I've been working on it off and on
aside from building gcc, it was mostly just getting the right CFLAGS to get it to build on solaris 10
TEttinger the only experience I have with sparc is with a bunch of ancient Sun Netra X1 servers that my brother's college threw out from the lab and he salvaged
orbus there was one minor patch needed in libuv
TEttinger (and they can't even run Solaris 10)
I ended up running Debian on them 02:48
learning how to netboot, crazy stuff for a CS novice
sold most of em for cheap on ebay.
orbus yup, there's lots of old sparc hardware on ebay 02:49
TEttinger it does sell for considerably more than old x86
orbus it probably cost 10x as much when new too 02:50
TEttinger I'm curious if it could actually be a solaris bug and not a sparc-related bug
gcc on solaris might have some differences?
orbus don't know - it looked like there had been some work on solaris 11
well
TEttinger I know there used to be differences between make on solaris and on GNU/Linux 02:51
orbus yeah, there's sun make and gnu make
sun make doesn't work at all
for lots of open source software 02:52
TEttinger haha, that's exactly what I gathered from just googling anything related to building on solaris :)
orbus as for gcc, solaris 10 has gcc 3.4.3 as an optional package
TEttinger hm, is that rather old? 02:53
orbus yeah, latest is 5.2.something
moar has some built-ins that were added in gcc 4
or rather needs some built-ins
that's why I had to build gcc
TEttinger November 4, 2004 02:54
wow, that sounds a lot older than solaris 10
orbus but there's some notes I saw in the gcc docs after I did the build that make me think maybe I got a bad build
well, gcc isn't their primary compiler
TEttinger yeah, there's the sun studio one?
orbus right
02:54 JimmyZ_ joined
orbus but that's $$ 02:55
TEttinger mm.
orbus anyway, there were notes about needing to use specific versions of some of the library dependencies otherwise Bad Things happen
so I might try again later
was just curious if there was some known breakage on sparc 02:56
JimmyZ_ we are fine to run on more cpus, though the libuv patch needs to go to the upstream .
orbus I figured
it's basically just they're using cfmakeraw for some tty stuff 02:57
which solaris 10 doesn't have
the equivalent tty settings that cfmakeraw sets are actually in the cfmakeraw man page on linux haha
other than that it was all CFLAGS 02:58
needed to set various defines to get it to build
and of course it doesn't work yet
JimmyZ_ didn't use solaris at all😊 03:02
orbus I use it at work
and it would be nice if perl6 worked there someday 03:03
so I've just been tinkering with it
JimmyZ_ yeah
orbus jvm works but it's slooooow
Hello, World takes like 5 seconds to run
JimmyZ_ only if moarvm builds on it
orbus right, well that's the first hurdle 03:04
it does build
and like, I can get moar to output the help message
but when I tried to build nqp it dumps core
so something's not right
I still suspect my gcc build
JimmyZ_ yeah
orbus newer versions of gcc need like libmpc and a couple other libraries 03:05
and there were notes on the gcc site about those mis-compiling with older versions of gcc
and I had just grabbed the latest ones 03:06
so they may be silently broken
JimmyZ_ Is it big endian?
orbus yes
I saw the bug note out on github
about big-endian brokenness
but I saw some stuff in the irc logs suggesting that's fixed now
would have to ask jnthn I guess? 03:07
that's one of the reasons I stopped by the channel
JimmyZ_ not sure whether moarvm works well enough on big endian or not 03:08
orbus yeah
well it's all good
hopefully it will work eventually, and if I can help make that happen all the better
JimmyZ_ but we do have some code in moarvm to support it.
orbus cool 03:09
I think I'll re-build gcc when I have time, see if that improves the situation
I might try an older version 03:10
I think anything over 4.0 will work
oh 03:12
hmm, yeah, think I will have to try this again 03:15
for posterity, the CFLAGS I needed to set for the build were -D_FILE_OFFSET_BITS=32 -D_POSIX_PTHREAD_SEMANTICS -DSUNOS_NO_IFADDRS -DAO_USE_PTHREAD_DEFS 03:18
notes I made: 03:19
_FILE_OFFSET_BITS=32 is needed for libelf.h, which doesn't support 64 bit offsets on 32 bit compiles
_POSIX_PTHREAD_SEMANTICS needed for readdir_r arguments
SUNOS_NO_IFADDRS needed for libuv because Solaris 10 doesn't have ifaddrs.h
AO_USE_PTHREAD_DEFS needed for libatomic_ops because we get missing symbols otherwise
I tried to do a 64 bit build and more stuff broke 03:20
didn't push forward on that
solaris supports both 32 and 64 bit, but most userland stuff is 32 bit
JimmyZ_ so how can they use more memory on 32 bit... 03:24
orbus I think most big software packages (oracle, etc.) tend to be 64 bit 03:26
it's just like ls, grep, all that stuff ships 32 bit
03:42 diakopter joined 05:22 orbusbak joined 06:57 FROGGS joined 07:02 kjs_ joined 07:30 xiaomiao joined 07:35 kjs_ joined 07:40 zakharyas joined 07:44 kjs_ joined 08:11 lizmat joined 08:28 Ven joined 08:48 dalek joined 08:51 Ven joined 08:52 kjs_ joined
dalek Heuristic branch merge: pushed 48 commits to MoarVM/even-moar-jit by bdw 09:03
arVM/even-moar-jit: 3571e61 | brrt++ | src/jit/ (2 files):
C90 compatibility
 09:09
09:21 lizmat_ joined 09:37 xiaomiao joined 09:51 dalek joined 10:08 Ven joined
timotimo i've heard the talk brrt++ is currently giving won't have a recording available afterwards?! :( 10:40
nwc10 I see no camera
10:40 lizmat joined
dalek arVM: b335fdf | paultcochrane++ | src/6model/reprs/CStruct.c:
Avoid possible divide-by-zero error

This merely requires that the `multiple_of` variable is initialised to 1 instead of to 0. This change avoids a potentially nasty issue and I don't believe changes actual behaviour.		 10:40
arVM: 61ebda6 | FROGGS++ | src/6model/reprs/CStruct.c:
Merge pull request #251 from paultcochrane/pr/avoid-divide-by-zero

Avoid possible divide-by-zero error
timotimo maybe i can lobby later speakers to run a screen recorder software that also grabs audio from the mic during the talk
like OBS multiplatform 10:41
dalek arVM/cpp3: bfcc4b6 | FROGGS++ | / (7 files):
re-apply patch to add C++/NativeCall support
 10:44
arVM/cpp3: e624e11 | FROGGS++ | / (9 files):
re-apply patch to add C++/NativeCall support
 10:54
FROGGS (forced update) 10:55
11:02 dalek joined 11:19 camelia joined 11:36 Ven joined 12:04 kjs_ joined 12:06 Ven joined 12:34 lizmat joined 12:57 vendethiel joined 13:08 lizmat joined 13:11 brrt joined, kjs_ joined 13:12 FROGGS joined
brrt good * 13:16
FROGGS hi brrt 13:17
brrt: how was your talk going?
brrt it went well, actually, people said i was nervous, which i was 13:18
but also that they understood what i was talking about
which is kind of nice :-)
are you not at yapc::eu then? 13:27
FROGGS no, I was unable to afford the time+money :o( 13:33
brrt well, it's september alright, so i can imagine that 13:39
FROGGS it is more that I can't leave my wife and kids at home for two workshops/conferences in a row 13:41
hoelzro has anyone worked with pmurias at all to keep the js branch building with the MoarVM cross compiler? I'm trying to merge js & master in NQP, and I'm encountering issues that I believe are related to serialization format changes 13:42
brrt not me, aactually 13:44
13:54 dalek joined 14:00 kjs_ joined 14:04 orbus joined 14:09 FROGGS_ joined 14:16 kjs_ joined
timotimo no clue :| 14:19
brrt \o timotimo 14:20
timotimo can has slides? :)
brrt oh, yes, where shall i put them? 14:21
timotimo when i read you'll answer the question "will perl6 ever be fast" at the end of your abstract, i thought you'd pull a total jnthn and whip out benchmarks that show real-life perl6 code that does a 5x over perl5 or something :D
i mean, we already do that for number-crunching with tight loops, but that's easy-peasy :P 14:22
i don't care where the slides land :)
brrt no, i haven't done that 14:28
i have no spot to place them actually.... maybe at github 14:29
nope, github doesn't allow it 14:30
timotimo google drive? 14:40
gimme a second
hm, yeah, google drive is probably good 14:41
then i can put it up on t.h8.lv for you
and i suppse there's also space on www.p6c.org; dunno if it serves user's "public_html" directories 14:47
even if not, you can upload it to hack and i'll pull it over onto t.h8.lv
14:52 rarara joined 14:53 kjs_ joined
timotimo cat says: peeeeeeeeeeeeeeeeeeeeeeeeeeeoäääääääääääääääääääääääääääääääääääääääääääääääääääääe bbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbmppppppppppppppppppppppppppppppppppppppppppppppppppppppppppppppppppppppppppppppppppppppppppppppppppppppppppppp 15:21
15:47 diakopter joined 15:52 kjs_ joined 16:02 kjs_ joined 16:40 kjs_ joined 17:15 virtualsue joined 18:02 xiaomiao joined 18:03 lizmat joined 20:30 colomon joined 20:55 TEttinger joined
timotimo o/ 21:06
21:10 virtualsue joined 21:40 colomon 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!