<Hagfish> https://r13y.com/ hmm, i don't think i've seen this result before: "1670 out of 1672 (99.88%) paths in the minimal installation image are reproducible!" <Hagfish>ruby2.7.3-nokogiri-1.11.1.drv and ruby2.7.3-racc-1.5.2.drv are the culprits <Hagfish>not sure i've seen them on the list either <Melg8>no, that's new, before that gcc-10.2 were on list <stikonas>maybe cirrus doesn't have CAP_SYS_ADMIN... <Hagfish>it's weird to think that a couple of ruby packages are more difficult than gcc, and the kernel, etc. <Hagfish>but that's maybe an encouraging sign for a project that doesn't have ruby on the critical path ***sm2n_ is now known as sm2n
***Server sets mode: +cnt
<ekaitz>i sent a PR to bootstrap-seeds with the RV64I port of hex0 it works returning the result to the stdout but I don't know why is not able to open a file <ekaitz>that's why is set to WIP, but the algorithm it is actually working <xentrac>congratulations! that's huge progress! <xentrac>that's usually how I diagnose problems like "why isn't this opening a file" <ekaitz>oh xentrac I think I have something... hmmm <ekaitz>looks like it's opening the input file for writing :D <ekaitz>now the funny thing is that I don't really know why is it doing that LOL <xentrac>well, probably whatever register has th esecond argument to open() has the wrong value in it <ekaitz>i'll dig on it later... meanwhile if anyone wants to take a look there's the link to the PR <ekaitz>also, I didn't bother to close the files I opened, is that an issue? <xentrac>Linux will close them for you when your process exits <xentrac>I have seen programming environments (Pascal on MS-DOS) where failing to close output files can result in the output buffers not getting flushed, but I don't think that's an issue here <ekaitz>yeah... but it's a good practice and all that <ekaitz>i saw the other hex0 files don't close them neither so :) <ekaitz>aiight so I'll gdb on this little boy later and see what's going on... thanks for the help! <xentrac>in this context I would say to keep unnecessary code to a strict minimum, because it's more stuff to audit <ekaitz>xentrac: the register was loaded wrong. Just as you predicted *xentrac debugs ekaitz's code by telepathy <ekaitz>there was no other option... openat takes the arg in a register ;) <ekaitz>thanks! and thanks for the help! <xentrac>delighted my "help" was actual help ;) <ekaitz>just being there in the other side means a lot <xentrac>it's super exciting to get things up on RISC-V! <ekaitz>its a better architecture for a better world ;) <xentrac>oriansj was appalled by the instruction encoding :) <ekaitz>xentrac: some instructions are pretty crazy, yeah <ekaitz>in some of them the bits of the immediate are split in small chuncks and scrambled like crazy <xentrac>yeah, the spec actually explains why this is <xentrac>in their first few prototypes the fanout of the immediate sign bit was the limiting factor in the speed of the whole design <xentrac>so they redesigned the instruction format to be able to start that process at an earlier pipeline stage <xentrac>oriansj described it, incorrectly I think, as a premature optimization <stikonas>fossy: by the way, perl 5.34 is now out. Although, if it's non-trivial to upgrade, we can just stay with 5.32.1 ***hl_ is now known as hl
<fossy>stikonas: ill do that soon. Merge 5.32 for now. Differences are probably marginal <stikonas>fossy: well, it's ready to go, just remove that double space <stikonas>so I have 2 small PRs that can be merged <stikonas>but I'm running out of ideas with that draft PR