01:33
sharkcz_ has joined #fedora-riscv
01:36
tg_ has joined #fedora-riscv
01:40
sharkcz has quit [Ping timeout: 264 seconds]
01:40
tg has quit [Ping timeout: 264 seconds]
01:40
sharkcz_ is now known as sharkcz
01:41
tg_ has quit [Quit: tg_]
01:42
tg has joined #fedora-riscv
02:51
jednorozec has quit [Read error: Connection reset by peer]
02:52
jednorozec has joined #fedora-riscv
04:06
pakcjo has joined #fedora-riscv
04:07
<
pakcjo >
Hi!, I think i just got the Fedora-Developer-37-20221130.n.0-nvme.raw.img image running in qemu, I didn't modify extlinux so I don't have terminal output, however it is listenning to ssh connections but the password for root doesn't seem to work, i tried riscv and fedora_rocks!
04:13
<
pakcjo >
oh got it, user: riscv, password fedora_rocks!
05:09
zsun has joined #fedora-riscv
06:05
zsun has quit [Remote host closed the connection]
07:18
davidlt has joined #fedora-riscv
07:19
davidlt has quit [Remote host closed the connection]
07:24
davidlt has joined #fedora-riscv
07:32
zsun has joined #fedora-riscv
08:19
<
rwmjones_ >
ghc qemu build finished in:
08:19
<
rwmjones_ >
real 538m7.201s
08:19
<
rwmjones_ >
user 2201m57.893s
08:19
<
rwmjones_ >
sys 288m14.085s
08:21
<
davidlt >
rwmjones_, check ghci
08:22
<
davidlt >
that's without -j1?
08:25
<
rwmjones_ >
-j16 I think
08:25
<
rwmjones_ >
and ... (drum roll)
08:26
<
rwmjones_ >
$ ghc --interactive
08:26
<
rwmjones_ >
ghci> 1 + 2
08:27
<
rwmjones_ >
the package I rebuilt is basically what's in ghc rawhide dist-git (except I have a local patch to disable LTO)
08:27
<
rwmjones_ >
I'm going to go for pandoc now
08:27
<
rwmjones_ >
I'll also close without merging those PRs
08:29
<
davidlt >
rwmjones_, you don't need to go for pandoc now
08:29
<
davidlt >
We can just do a proper build in Koji
08:32
<
rwmjones_ >
sure, I just want to check there's nothing else and actually build it
08:33
<
rwmjones_ >
but also please do a build in koji if you want
08:33
<
davidlt >
rwmjones_, yeah, but hash probably changed and you would need to rebuild pretty much all packages
08:33
<
rwmjones_ >
I'll find out soon
08:36
<
davidlt >
I will boot Pioneer in a minute and start looking
08:45
<
davidlt >
updated gitea to 1.21.5
08:47
<
rwmjones_ >
the hashes didn't change
08:48
<
rwmjones_ >
plus confirmed I can rebuild those packages without the no-test patches
08:50
<
rwmjones_ >
now rebuilding ghc-pandoc-types
08:50
<
rwmjones_ >
building, not rebuilding
08:53
<
davidlt >
I removed all the changes from ghc-rpm-macros for now (imported, tagged)
08:54
<
davidlt >
checking ghc actual diff
08:58
<
davidlt >
rwmjones_, just to confirm, after looking at the diff the current upstream dist-git seems to be OK for riscv64 now?
08:58
<
davidlt >
Pioneer connected to Kojiii
09:02
BryceII has quit [Ping timeout: 256 seconds]
09:04
<
davidlt >
all these changes aren't built in Fedora 40 yet, thus there is no new NVR
09:04
<
davidlt >
otherwise I launched a build
09:05
<
davidlt >
We should look at 9.6.X too as that's a new default for Fedora 40
09:07
zsun has quit [Quit: Leaving.]
09:08
<
rwmjones_ >
davidlt: current ghc rawhide dist-git is working fine
09:10
<
rwmjones_ >
9.6.x isn't in fedora yet though ..?
09:11
<
rwmjones_ >
anyway I think we need 9.4 built before we could do 9.6
09:11
<
davidlt >
9.6 and 9.8 should be in (we pretty much have all stable GHC versions), but 9.6 is not yet default.
09:11
<
rwmjones_ >
I don't understand, "in" where?
09:11
<
davidlt >
There is approved change request for 9.6 to be default.
09:11
<
rwmjones_ >
ok I don't see it in dist-git
09:11
<
davidlt >
Fedora incl. all relevant GHC versions (similar to Python)
09:11
<
rwmjones_ >
is there another package apart from "ghc"?
09:12
<
rwmjones_ >
ok I didn't see those til now
09:12
<
rwmjones_ >
weird way to package stuff ..
09:12
<
davidlt >
Not really, it's done with Python too.
09:13
<
davidlt >
Basically we change default based on "LTS Haskell"
09:13
<
rwmjones_ >
how do the ghc libraries work? are there ghc9.x-<lib> packages too?
09:13
<
davidlt >
LTS 22 was announced with 9.6, and 9.8 is now in nightly.
09:13
<
davidlt >
No, those don't work IIUC.
09:13
<
davidlt >
But GHC itself is available.
09:14
<
rwmjones_ >
afaict you need libraries (hundreds of em!) to be able to do anything useful in haskell
09:14
<
rwmjones_ >
ghc-pandoc-types built
09:14
<
davidlt >
Yeah, but having a working compiler is good once you need to switch :)
09:14
<
davidlt >
LTS Haskell stuff is always a few compiler versions behind.
09:16
<
rwmjones_ >
ok I've got another couple of pages of deps to try before I can start pandoc
09:27
<
davidlt >
rwmjones_, it's cooking now
09:28
<
davidlt >
so far load averages are low, ~4 but it's -j 64
09:45
iooi_ has joined #fedora-riscv
09:47
iooi__ has joined #fedora-riscv
09:47
iooi has quit [Ping timeout: 264 seconds]
09:47
iooi__ is now known as iooi
09:50
iooi_ has quit [Ping timeout: 252 seconds]
10:18
<
davidlt >
GHC loves single-thread perf
10:18
<
davidlt >
I have htop running to see what happens, and so far it's mostly <4 cores
10:40
<
rwmjones_ >
ghc-hslua-module-text seems to depend on itself!
10:49
<
davidlt >
yeah, there are a few packages that need cycles nuked to boostrap
10:49
<
davidlt >
these days there is usually bootstrap conditional in the spec
10:54
<
rwmjones_ >
I don't see any bootstrap flag in this one
11:13
<
davidlt >
it might not be in this one
12:09
davidlt has quit [Ping timeout: 255 seconds]
13:01
zsun has joined #fedora-riscv
13:03
zsun has quit [Client Quit]
13:06
davidlt has joined #fedora-riscv
13:07
<
davidlt >
looking at GHC build I really want to see JH8100-series boards
13:08
<
davidlt >
4+1+1 cores seems like a nice balance
13:09
<
davidlt >
I am linking designs like big.LITTLE, P-cores + E-cores [+ LP-E cores ] are looking more interesting
13:10
<
davidlt >
I hope Intel delivers 8 + 32 (long-term, Arrow Lake earliest) with AVX10.2 on all cores
13:55
zsun has joined #fedora-riscv
14:03
pakcjo has quit [Ping timeout: 268 seconds]
14:30
zsun has quit [Quit: Leaving.]
14:44
<
rwmjones_ >
top - 14:44:41 up 18 days, 23:59, 32 users, load average: 18.15, 13.50, 7.46
14:44
<
rwmjones_ >
that's on my 4 core vf2 ..
14:50
<
davidlt >
what are you doing?
14:50
<
davidlt >
those loads seems too high
14:55
pakcjo has joined #fedora-riscv
15:01
<
rwmjones_ >
building about 30 packages in parallel
15:08
<
davidlt >
am, is that smart?
15:16
<
rwmjones_ >
well, it's interesting
15:47
<
rwmjones_ >
do you know what the llc-14 process is? llvm C compiler?
15:49
<
rwmjones_ >
oh, it just finished (after 1 hour of running)
15:53
<
davidlt >
Yes, that's from LLVM
16:07
<
rwmjones_ >
ok now I'm tackling that ghc lua stuff
16:07
<
rwmjones_ >
it's all one package as far as I can tell!
16:10
<
rwmjones_ >
branching is happening
16:10
<
rwmjones_ >
also koji is down :-(
16:10
<
rwmjones_ >
(I mean fedora koji)
16:12
<
rwmjones_ >
ok the above about hslua is wrong .. repoquery was giving me wrong answers for some reason
16:12
<
rwmjones_ >
in fact there's ghc-hslua -> ghc-hslua-module-text
16:26
davidlt has quit [Ping timeout: 260 seconds]
17:02
davidlt has joined #fedora-riscv
17:10
<
rwmjones_ >
down to only 5 missing pandoc deps
17:26
davidlt has quit [Ping timeout: 255 seconds]
17:43
davidlt has joined #fedora-riscv
17:45
<
davidlt >
rwmjones_, it will work
17:57
pakcjo has left #fedora-riscv [#fedora-riscv]
18:20
<
rwmjones_ >
yeah I'm down to just one missing build dep, which I'm building now
18:29
<
rwmjones_ >
ok pandoc is building!
18:30
<
davidlt >
it will work :)
18:41
<
davidlt >
rwmjones_, ghc is writing out RPMs
18:41
<
davidlt >
all 64 cores 100% :)
18:41
<
davidlt >
That means it will take some time
19:02
<
rwmjones_ >
I guess pandoc's gonna take 12 hours or so
19:02
<
rwmjones_ >
it takes 1.5 hours on koji x86-64
20:58
davidlt has quit [Ping timeout: 255 seconds]