github.com/moarvm/moarvm | IRC logs at colabti.org/irclogger/irclogger_logs/moarvm
Set by AlexDaniel on 12 June 2018.
00:03 Altai-man_ joined 00:06 sena_kun left 00:43 lucasb left 00:58 Altai-man_ left 01:11 moon-child left 01:16 moon-child joined 03:42 colomon joined 04:37 colomon left 07:09 reportable6 left 07:11 reportable6 joined 07:19 domidumont joined 08:06 zakharyas joined
Geth MoarVM: af34daa429 | (Stefan Seifert)++ | 10 files
Enable warnings about implicit fallthrough in switch statements on GCC

A break; statement can be easily forgotten resulting in obscure bugs. Luckily gcc can warn us about them if told to do so, which we now do. Of course in some cases a fallthrough is done on purpose. These cases can be marked with the new MVM_FALLTHROUGH macro.		 08:25
08:33 domidumont left 08:39 travis-ci joined
travis-ci MoarVM build failed. Stefan Seifert 'Enable warnings about implicit fallthrough in switch statements on GCC 08:39
travis-ci.org/MoarVM/MoarVM/builds/618084959 github.com/MoarVM/MoarVM/compare/8...34daa429f8
08:39 travis-ci left
nwc10 nine: gcc: error: unrecognized command line option ‘-Wimplicit-fallthrough=3’ 08:49
doesn't work on "my" machine, for the value of "my" that has an old gcc.
Not sure how MoarVM is supposed to probe for this sort of stuff.
but cool find - I wasn't aware of that flag. It might be about to be added to the $ork code
nine Apparently it was introduced with gcc 7 08:50
I can probably hack in a version check for gcc. Won't be pretty though 08:51
An alternative would be to enable -Wextra instead which includes the same check. But that'd require fixing boat loads of new warnings... 09:00
Which may be worth it though
09:11 patrickb joined
patrickb nine: This commit also adds a probe for a gcc warning feature. Maybe it helps as inspiration. github.com/MoarVM/MoarVM/commit/23...34ec239568 09:12
09:32 robertle joined
nine nqp: my $t := nqp::newtype(nqp::knowhow(), "MultiDimArray"); nqp::composetype($t, nqp::hash("array", nqp::hash("dimensions", 2))); nqp::writeint(nqp::create($t), 0, 0, 0) 09:48
camelia (signal SEGV)
nine Discovered through -Wextra :)
09:51 domidumont joined 10:16 sena_kun joined
nine Oooh: src/strings/gb2312.c:15:41: warning: comparison is always true due to limited range of data type [-Wtype-limits] 10:39
Geth MoarVM/master: 11 commits pushed by (Stefan Seifert)++
review: github.com/MoarVM/MoarVM/compare/a...8ed2b03039		 11:21
nine What's left is loads of warnings about comparison of integer expressions of different signedness 11:22
11:28 domidumont left
nwc10 $ork code had a load of them. I systematically removed them, but it sort of felt a bit like makework 11:31
dogbert17 hmm, can gcc 5.40 be considered old ? 11:33
nwc10 dogbert17: no no no, it's from the future. Given access.redhat.com/solutions/19458 11:36
11:36 travis-ci joined
travis-ci MoarVM build failed. Stefan Seifert 'Avoid comparisons that are always true 11:36
travis-ci.org/MoarVM/MoarVM/builds/618148473 github.com/MoarVM/MoarVM/compare/a...8ed2b03039
11:36 travis-ci left
nwc10 src/gen/config.h:19:25: error: ISO C90 forbids mixed declarations and code [-Werror=declaration-after-statement] 11:37
#define MVM_FALLTHROUGH __attribute__ ((fallthrough));
that's what I see
dogbert17 yeah, that's what I see as well 11:38
11:39 colomon joined
dogbert17 ah, ofc it was version 5.4.0 :) 11:39
On my RPi4 I have gcc 8.3.0 11:40
nwc10 total digression, but it was pleasing that I could put the SD card from the RPi4 (in an adaptor, of course) into a very early RPi B, and it (also) boots happily on it. 11:41
dogbert17 that's cool 11:43
nine That....feels more like a compiler bug 11:44
dogbert17 building Raku on RPi 4 is still a relatively slow process, the parse step takes ~250s on that system compared to 40s on my normal machine
nine dogbert17: solution: build on a faster machine 11:47
dogbert17 Crosscompiling you mean 11:48
nine moarvm files are the same on every architecture. No reason to compile them on an underpowered machine
dogbert17 aha 11:49
btw, my RPi 4 just spit out the following warning:
src/vm/moar/ops/perl6_ops.c:25:1: warning: missing initializer for field ‘is_interned’ of ‘MVMCallsite’ {aka ‘struct MVMCallsite’} [-Wmissing-field-initializers]
static MVMCallsite no_arg_callsite = { NULL, 0, 0, 0, 0 };
^~~~~~
In file included from /home/pi/repos/rakudo/nqp/MoarVM/../../install/include/moar/moar.h:102, 11:50
from src/vm/moar/ops/perl6_ops.c:2:
is that something to get worried about?
nine no
dogbert17 phew
11:54 MasterDuke joined 12:04 Altai-man_ joined 12:07 sena_kun left 12:21 zakharyas left
MasterDuke nine: clang throws a bunch of different warnings 12:25
nine MasterDuke: I guess that's good :) 12:27
MasterDuke gist.github.com/MasterDuke17/f46a4...c10b484b50 some of them (plus some non-int comparison ones from gcc) 12:28
gist updated with a couple more clang warnings 12:33
nine What does if (result < -1 || result >= 15) actually do wehn result is an unsigned integer? 12:48
12:49 lucasb joined 12:55 domidumont joined
nine We are really sloppy with our treatment of signness 13:10
13:15 AlexDaniel` left
Geth MoarVM/master: 6 commits pushed by (Stefan Seifert)++ 13:30
nine VMArray can deal with up to 2^64 elements while the ops to access them can only address 2^63 13:33
MasterDuke heh 13:36
13:44 travis-ci joined
travis-ci MoarVM build failed. Stefan Seifert 'Fix signedness errors in IO code' 13:44
travis-ci.org/MoarVM/MoarVM/builds/618198309 github.com/MoarVM/MoarVM/compare/c...a1c25a5d96
13:44 travis-ci left
Geth MoarVM: 9bce55092a | (Stefan Seifert)++ | src/6model/reprs/VMArray.c
Fix signedness errors in VMArray
 13:45
MoarVM: 5d5b86e30c | (Stefan Seifert)++ | src/jit/compile.h
Use unsigned values for counting things in JIT graph
lizmat nine++ # fixing sloppiness 13:47
nine This is funny: a P6opaque may have 2^16 attributes, but delegates for positional or associative things or box targets must be within the first 2^15 and just to be on the safe side we use a 32 bit unsigned for the name to index map 13:56
Of course we then only use a signed 16 bit integer to actually iterator over the name to index map ;)
14:00 travis-ci joined
travis-ci MoarVM build failed. Stefan Seifert 'Use unsigned values for counting things in JIT graph' 14:00
travis-ci.org/MoarVM/MoarVM/builds/618204604 github.com/MoarVM/MoarVM/compare/9...5b86e30c55
14:00 travis-ci left
Geth MoarVM: 85f5604e79 | (Stefan Seifert)++ | src/6model/reprs/P6opaque.c
Fix potential endless loop with objects with > 2^15 attributes
 14:03
14:05 sena_kun joined
Geth MoarVM: a0d545b134 | (Stefan Seifert)++ | build/setup.pm
Try to fix build on Travis

Older GCC versions throw "error: ISO C90 forbids mixed declarations and code" when there actually aren't any declarations following (we're in the middle of a switch statement).		 14:05
14:06 Altai-man_ left 14:15 zakharyas joined
Geth MoarVM: 9a59fdc9a0 | (Stefan Seifert)++ | src/jit/stub.c
Fix build without JIT support

Mismatched type due to the superfluous const		 14:17
14:18 travis-ci joined
travis-ci MoarVM build failed. Stefan Seifert 'Fix potential endless loop with objects with > 2^15 attributes' 14:18
travis-ci.org/MoarVM/MoarVM/builds/618211858 github.com/MoarVM/MoarVM/compare/5...f5604e79b0
14:18 travis-ci left 14:29 travis-ci joined
travis-ci MoarVM build failed. Stefan Seifert 'Try to fix build on Travis 14:29
travis-ci.org/MoarVM/MoarVM/builds/618212911 github.com/MoarVM/MoarVM/compare/8...d545b13485
14:29 travis-ci left
Geth MoarVM: MasterDuke17++ created pull request #1216:
Fix 'may be used uninitialized' warning		 14:30
MoarVM: b6327c6034 | (Daniel Green)++ | src/main.c
Fix 'may be used uninitialized' warning
 14:31
MoarVM: ff73cf1641 | niner++ (committed using GitHub Web editor) | src/main.c
Merge pull request #1216 from MasterDuke17/fix_telemeh_use_uninitialized_warning

Fix 'may be used uninitialized' warning
AlexDaniel samcv: ping? 14:44
14:47 colomon left 14:49 travis-ci joined
travis-ci MoarVM build passed. Stefan Seifert 'Fix build without JIT support 14:49
travis-ci.org/MoarVM/MoarVM/builds/618217322 github.com/MoarVM/MoarVM/compare/a...59fdc9a0b0
14:49 travis-ci left 14:52 zakharyas left 14:59 AlexDaniel` joined 15:00 robertle left 15:01 zakharyas joined, travis-ci joined
travis-ci MoarVM build passed. niner 'Merge pull request #1216 from MasterDuke17/fix_telemeh_use_uninitialized_warning 15:01
travis-ci.org/MoarVM/MoarVM/builds/618223645 github.com/MoarVM/MoarVM/compare/9...73cf1641ee
15:01 travis-ci left
dogbert17 nine++ I can now build again 15:03
Geth MoarVM: 0150c1d828 | (Stefan Seifert)++ | 4 files
Fix a couple of variables to match bytecode format definition
 15:07
MoarVM: 29acd43f0c | (Stefan Seifert)++ | 5 files
Fix mixed signedness comparisons ins SC handling
MoarVM: a06a0edf37 | (Stefan Seifert)++ | 2 files
Use unsigned values in serialization reader/writer to match bytecode format
dogbert17 what does this mean? src/core/interp.c:104:6: warning: variable ‘val’ might be clobbered by ‘longjmp’ or ‘vfork’ [-Wclobbered] 15:11
15:16 zakharyas left 15:38 zakharyas joined 16:04 Altai-man_ joined 16:06 sena_kun left
nwc10 dogbert17: you read this? www.raspberrypi.org/blog/thermal-t...erry-pi-4/ -- vertical is coolest 16:18
16:22 patrickb left 16:32 domidumont left
Kaiepi is github.com/MoarVM/MoarVM/pull/1214 good to go now nine? 17:41
18:05 sena_kun joined 18:07 Altai-man_ left 18:25 Guest38485 left
Kaiepi what architectures are supported by moarvm? 18:41
18:41 Guest38485 joined 19:13 MasterDuke left 19:23 zakharyas left 20:01 Ven`` joined 20:04 Altai-man_ joined 20:06 sena_kun left
lucasb no tarball yet... 20:08
20:13 Ven`` left
lizmat samcv is dealing with a lot at $work, is my understanding :-( 20:14
nine Can someone else generate/upload the tar ball? 20:16
20:38 MasterDuke joined
MasterDuke nine: while you're on a roll, gist.github.com/MasterDuke17/f46a4...c10b484b50 is updated with some new warnings that just popped up when i recompiled rakuod 20:39
Geth MoarVM: 4f4c304baf | (Stefan Seifert)++ | 21 files
Fix signedness mismatches with bytecode definition in spesh
jnthn nine: Yes, if they're set up to be able to sign stuff. Check out the release tag, do 6, 7, 14, 15, 16 20:47
(Not instructing you to do it, just saying how it's done.)
I need to get set up to sign stuff; I'll get to it when things calm down a little
20:49 travis-ci joined
travis-ci MoarVM build failed. Stefan Seifert 'Fix signedness mismatches with bytecode definition in spesh' 20:49
travis-ci.org/MoarVM/MoarVM/builds/618350186 github.com/MoarVM/MoarVM/compare/a...4c304bafc1
20:49 travis-ci left
MasterDuke nine: here's a backtrace from the new segv when building nqp gist.github.com/MasterDuke17/2e019...cd67c95517 20:54
disabling spesh makes the segv go away 20:55
nine Of course, the one time I accidentally pushed before running the spectest it explodes 20:56
At least I got a fix 20:57
MasterDuke weird though, doesn't seem to be in code you touched, at least not directly
nine It's about spesh stats, and I touched spesh stats
AlexDaniel nine: more like no because it's a hassle. We don't have any team key or anything like that, so a different release manager means a different key, and then those who check it properly will need to adapt their stuff
the real problem in this case, I think, is that there's not enough automation for moarvm 20:58
that's not the first time some tiny (but annoying) mistake happened during the release process, just because it's not automated enough 20:59
MasterDuke nine: ahh, here? github.com/MoarVM/MoarVM/commit/4f...15cba2L501
Geth MoarVM: 61205646b0 | (Stefan Seifert)++ | src/spesh/stats.c
Fixup "Fix signedness mismatches with bytecode definition in spesh"

Missed to places where I needed to replace the variable.
21:17 travis-ci joined
travis-ci MoarVM build passed. Stefan Seifert 'Fixup "Fix signedness mismatches with bytecode definition in spesh" 21:17
travis-ci.org/MoarVM/MoarVM/builds/618355686 github.com/MoarVM/MoarVM/compare/4...205646b099
21:17 travis-ci left
Geth MoarVM: d4f20a191c | (Stefan Seifert)++ | 10 files
Fix signedness mismatches in string processing code
 21:32
22:05 sena_kun joined 22:06 Altai-man_ left 23:00 jjatria joined
samcv sorry about lack of tarballs. i released then went a bit away from IRC for a while 23:25
AlexDaniel thanks! 23:27
samcv: can you tweak the wikipedia article too? :)
lizmat afk& 23:29

Powered by Cro and the Raku® Programming Language.

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

Thank you!