[Coke] ugh. 00:22
Wish the protest rally was not scheduled same day as the release I am signed up for.
lizmat the rally has priority 00:33
simple as that
[Coke] I'll do both, I'll just be drinking a lot of coffee that day! 00:35
04:32 rakkable left 04:33 rakkable joined
disbot2 <librasteve> on .1e0, this ofc is a creature of the FPU, whereas 0.1 résides in the CPU … so even in hand tuned asm, you still need to get the Rat values from the CPU register stack to the FPU to convert to num32 … and I suppose in MOARVM this is done via main memory and a cache flush or two … 08:14
<librasteve> i’ve been using FatRat a lot for range and precision beyond num64 so that’s a nice way to avoid these traversals in userland … Num remains a great way to get more performance for most repetitive calculations… but I would say write your FPU code in a block and be careful about conversion is good advice and true for all languages 08:22
<librasteve> sorry … you guys probably all know this 08:23
timo all you need to do to do floating point math with a value is to load it into an sse register nowadays 09:01
09:54 finanalyst joined 10:40 japhb left 10:44 japhb joined 10:48 Geth left, Geth joined 10:49 Geth joined 10:52 japhb joined, finanalyst joined, rakkable joined, [Coke] joined, Pixi joined, leont joined, coleman joined, refactus joined, benchable6__ joined, coverable6__ joined, sourceable6__ joined, linkable6__ joined, tellable6__ joined, vrurg joined, ShimmerFairy joined, disbot2 joined, kjp joined, nine joined, bisectable6 joined, notable6 joined, committable6 joined, quotable6 joined, greppable6 joined, releasable6 joined, shareable6 joined, unicodable6 joined, bloatable6 joined, nativecallable6 joined, evalable6 joined, patrickb joined, sivoais joined, Voldenet joined, rba joined, summerisle joined, JRaspass joined, [Tux] joined, sjn joined, jjatria joined, lucs joined, djinni` joined, ab5tract joined, SmokeMachine joined, gfldex joined, timo joined, zostay joined, tbrowder joined, samebchase joined, camelia joined, ilogger2 joined, jdv joined 13:27 sp1983 joined
sp1983 o/ 13:27
my code randomly segfaulted (on fresh Rakudo?) with coredump - github.com/rakudo/rakudo/issues/6095 13:28
if someone would take a look I'd appreciate , thanks
Other users of Sparrow6 report this two from time to time, the bug is sporadic and hard to reproduce ... 13:29
two -> too 13:30
[Coke] Do we want to try to migrate everyone on CPAN over to fez? 13:33
(not forcefully, more like write up a paragraph of rationale, and then open a bunch of tickets with that) 13:34
Also, I just found an old dist of mine that was failing because I had a .pm6 file lying around.
also, can we try another run of the blin replacement this week? I have to kick off a blin run today at the latest. 13:35
(in addition to re-trying the potential replacement)
13:37 sp1983 left
[Coke] do we have any pending changes in moarvm/nqp that need to be pulled into rakudo/main for testing? 13:42
lizmat afaik main in MoarVM and nqp are reflected in main on rakudo 14:03
[Coke] cool, thanks, lizmat 14:05
disbot2 <librasteve> timo: TIL - thanks for the "news" ;-) 14:20
14:46 finanalyst left
timo separate FPUs haven't been a thing in x86 for a long long time :) 14:46
lizmat yeah, I was thinking the same thing :-) 14:47
timo but I couldn't tell you with confidence when exactly doing FPU stuff became less complicated than having to push to the FPU's value stack or whatever it used to have 14:48
I guess with the introduction of MMX? maybe there was something else before MMX that did floating point work with a separate set of registers
[Coke] is #6095 a blocker for .03? (looks like it was found in the wild in .02?) 14:54
15:00 librasteve_ joined
disbot2 <librasteve> well, back in the day I worked for Weitek -so I have a deep knowledge of ancient history (when the FPU didn't even fit on the same chip as the CPU) ;-) 15:04
lizmat yeah, it was an optional add-on for quite some time: if you didn't have it, floating point ops would be *sloooow* 15:14
timo in the embedded world you still find processors where you have to use "soft floats", i.e. you ask the compiler to spit out implementations of floating point math based on integer arithmetic 15:35
disbot2 <melezhik.> To make it clear #6095 relates to the latest (2026.02) Rakudo version right ? 15:40
<melezhik.> Coke - ooc what is blin replacement? ) 15:54
[Coke] ... didn't you write it? 16:15
I should have said "potential"
timo @melezhik., can I take a look at your core dump(s)? 16:17
I would probably also need the libmoar.so at least to be able to see much of anything
do you happen to have a date&time for the dmesg output? dmesg -H shows a human-readable date&timestamp instead of seconds-since-boot 16:18
oh, also: does this happen quickly after starting to run that? 16:20
disbot2 <melezhik.> It looks like happen during start at “use Sparrow6::DSL” statement 16:33
<melezhik.> GitHub does allow attach .so files but basically this then one that comes with distro 16:43
<melezhik.> Added dmesg -H info to the ticket
<melezhik.> Coke as for the brownie - I am all for it, we just need to plan . As I am in a bit tight time budget right now 16:44
<melezhik.> Attached libmoar.so to the ticket 16:46
[Coke] no pressure. blin workingb for now 16:59
timo ok, you have an rpm based distro of some kind there, can you tell me which exact one it is so i can best reproduce it locally? 17:00
do you have a sparrowfile for me that I can try to reproduce the issue with? 17:11
disbot2 <melezhik.> It’s installed from nxadm-packages repo
timo right, i have an opensuse leap 16 container here where i ran `zef install Sparrow6` right now 17:12
disbot2 <melezhik.> It’s not just one sparrowfile, it’s sparky scenario launching qemu, could be a bit more complicated , but I can point you a repo with readme
<melezhik.> But I guess maybe you can golf it out that way
timo i have an s6 binary now, which isn't the same thing as .sparrowdo/sparrowrun.sh, but it doesn't necessarily have to be
disbot2 <melezhik.> As far as I can get you reproduces via just raku -MSparrow6::DSL -e 1 17:14
<melezhik.> Something
timo 4be1c9a3eb1e:/ # raku -MSparrow6::DSL sparrowfile
WARNINGS for //sparrowfile: 17:15
Useless use of constant integer 1 in sink context (line 1)
no crash on my end
before we do anything else, can you back up the entire suite of precompiled files and such? becausea corruption in those files can perhaps cause segfaults and we may want to try reinstalling all your modules to see if that makes the error go away 17:16
disbot2 <melezhik.> Yeah . It starts under some circumstances .
timo (but if we just clean stuff out and reinstall the modules, and it does fix the issue, we still need to hunt down the exact part that went bad)
disbot2 <melezhik.> Yeah . The problem was is . The issue happened only once on my VM, but has not repeated since then . And keeps repeating on another user VM, which I don’t control 17:17
timo what surprises me the most tbh is that it outputs a partial stacktrace, which is not very normal for a segfault
disbot2 <melezhik.> Maybe if course he may something messed up with his setup 17:18
timo oh, you can't just reproduce it at will right now?
that's tricky
disbot2 <melezhik.> No. It’s not reproduced on my machine
<melezhik.> It happened few days ago on 2 others users machines and keeps happening today 17:19
<melezhik.> I awaiting for his new reports
<melezhik.> However problem is we run the same scenario few times, and the issue seems to happen randomly 17:20
<melezhik.> So with all being said lets wait for new reports … 17:23
timo you said you have a core dump though? 17:24
it could obviously contain secrets and such, so it makes sense not to share it publicly on the channel here 17:25
disbot2 <melezhik.> No. I don’t have core dump . At least on my machine . I guess core dump needs to be enabled on OS level ? 17:26
timo if it says "(core dumped)" then it has written out a core dump, but where it goes is determined by your kernel's settings
namely /proc/sys/kernel/core_pattern 17:27
disbot2 <melezhik.> Ah . Ok. Will take a look
timo on a typical linux with journald and systemd you would probably find it with `coredumpctl list` if it's not fallen off the end yet
disbot2 <melezhik.> Ok 17:30
timo and with `coredumpctl info` you can get backtraces of all threads for free, and `coredumpctl gdb` would drop you into gdb with the core dump loaded 17:31
disbot2 <melezhik.> I have this
<melezhik.> gist.github.com/melezhik/4f775716e...abd0ee3bbf
<melezhik.> Let me grab files , where to put them ?
<melezhik.> Updated gist 17:33
timo "truncated"? :(
disbot2 <melezhik.> The one of the core dump seems not relevant
timo that looks to be a crash of the process [...]share/perl6/site/bin/tom.raku install 17:34
disbot2 <melezhik.> Yes. But this is not what the original ticket about 17:35
<melezhik.> How can I see details on another core dump ?
timo `coredumpctl info rakudo` should show info for both 17:36
it takes "matches" at the end to figure out which you want, but "info" will output any that match if you give it matches, and only the very latest of all if you don't
disbot2 <melezhik.> Looks I need pass pid as parameter 17:37
timo that should also work, yeah
disbot2 <melezhik.> Here we are , updated gist
<melezhik.> This time it’s what original ticket is about 17:38
timo hm, ok, not all hope is lost, that uppermost frame is maybe just a JIT frame and if we're in GDB we can just take the frame pointer to get back to a real stack trace? hopefully?
now, how to share a file over the internet ... one of the most difficult tasks of modern day people 17:39
we can try using this sendfiles.dev/ for a peer-to-peer file transfer, you would DM me the link it generates, and we would verify the sha256 afterwards to be sure 17:40
disbot2 <melezhik.> So, you need me to upload coredump there , right ?
timo yeah
well, it's not so much an "upload"
it just sets up a direct connection between our computers that the file can go through 17:41
wait
i forgot you're on discord
i don't think our bot lets DMs through to irc 17:42
but i'm also on discord, hold on
disbot2 <melezhik.> Ok
<tpaul64_18694> sup
timo ^- that's me
disbot2 <melezhik.> Ok
timo haha, I spot a wild Voldenet in another channel 17:51
Voldenet coincidences happen ;) 18:31
timo wee 18:33
Voldenet I've been there since reporting one bug 15 years ago :) 18:36
timo one bug in 15 years that's a good track record for software 18:39
releasable6 Next release in ≈4 days and ≈23 hours. There are no known blockers. Please log your changes in the ChangeLog: github.com/rakudo/rakudo/wiki/ChangeLog-Draft 19:00
timo from a bit more context of the segfault line in the dmesg I saw that the instruction pointer landed right in the section of libmoar.so where names of hangul syllables live 19:01
librasteve_ rakudoweekly.blog/2026/03/23/2026-...-berliner/ 19:21
timo librasteve_: some code inside of < > angle brackets got eaten 19:24
disbot2 <librasteve> ... 19:25
<librasteve> err I think this is referring to the angle backets themselves - can you provide a little context? 19:26
timo > dub ; #4
disbot2 <librasteve> ta
timo it looks like dub is being called without arguments
disbot2 <librasteve> yeah and the dd 19:27
<librasteve> thanks! 19:30
<librasteve> (I''l revert to Rainbow / Hilite next time ... but have patched with Wordpress perl hilights for now 19:31
<librasteve> )
23:15 kjp left