IRC channel logs

<wingo>ACTION cannot remember why psyntax labels are strings instead of integers

<sneek>wingo, you have 1 message!

<sneek>wingo, k4r4b3y says: that I would appreciate help on https://github.com/wingo/tekuti/issues/4

<ray1729>wingo: I was able to help k4r4b3y

<dpk>wingo: i was wondering that myself …

<morenonatural>hey, y'all can/should I serialize a vhash table into a file? Should I use a different key-value storage that I can write to/read from a file?

<wingo>morenonatural: easiest is an alist tbh

<wingo>for serialization

<morenonatural>how do I serialize an alist? sxml->xml complained on '(("a" .1))

<morenonatural>wingo, almost forgot: thanks

<ekaitz>morenonatural: you can just `write` it

<ekaitz>morenonatural: if you need xml for any reason, you should make proper sxml

<wingo>object->string if you need to embed a datum in sxml as text

<ekaitz>wingo: i'm trying to make fix the lightening riscv port, i have some segfault that i don't know how to debug, any suggestion?

<ekaitz>wingo: I just build guile with the jit activated and when it starts it segfaults, but I if I gdb on it, i don't have any reference to where it happened

<wingo>ekaitz: have you tried bisecting to find the function whose JIT causes the error?

<ekaitz>wingo: bisecting how? you mean git bisect?

<wingo>GUILE_JIT_STOP_AFTER=42 guile

<ekaitz>hm!

<ekaitz>i didn't try that

<wingo>also turn up GUILE_JIT_LOG, etc

<ekaitz>and how can I find the proper one?

<ekaitz>the LOG i tried, and it made sense

<wingo>well. you run your program with GUILE_JIT_STOP_AFTER=10000000000 and then if it fails you try half that, etc

<morenonatural>ekaitz, what function should I use to read it back to an object?

<ekaitz>wingo: oh manual bisection

<ekaitz>morenonatural: i suggested `read` and `write`

<ekaitz>morenonatural: those are the basic functions to write and read scheme data/code

<morenonatural>ekaitz, 🙌🏾

<morenonatural>thx

<ekaitz>morenonatural: no problem

<wingo>ekaitz: yeah manual bisection

<ekaitz>wingo: GUILE_JIT_STOP_AFTER=8 is ok and 9 is broken

<wingo>great :) now you know the broken code

<ekaitz>but how? haha i don't know what it represents very well

<wingo>try GUILE_JIT_LOG=3, then GUILE_JIT_PAUSE_WHEN_STOPPING=1 to attach with gdb

<ekaitz>haha

<wingo>then disassemble that code range

<ekaitz>hmmm

<wingo>goal is to figure out which op caused bad code, or maybe guile is using registers in an unexpected way, etc

<ekaitz>I connected with gdb

<ekaitz>hmm

<ekaitz>so i stop at 8, but then it doesn't fail so how can i find the issue with gdb?

<ekaitz>should I stop in 9 and pause when stopping?

<ekaitz>oh i think i'm starting to figure out the idea

<ekaitz>lol this is hard

<ekaitz>jit: entering mcode: 0x3fbb9f3820

<ekaitz>that segfaults

<ekaitz>it segfaults right after that, better said

<ekaitz>jit: mcode: 0x3faca0e820,+4384 <-- then i try to disassemble that and 'Cannot access memory at address 0x3faca0e820'

<ekaitz>hm!

<ekaitz>oh i was doing it wrong

<ekaitz>okay so i think i know what's wrong, there's a huge value in t2

<ekaitz>t2 0x7fefa97c00000000

<ekaitz>and it's trying to `jr t2`

<ekaitz>wingo: now I know how to debug this, thanks for explaining, it wasn't obvious without the explanation

<ekaitz>wingo: we'll see if I manage to fix it

<wingo>ekaitz: good luck and do poke me whenever you like, will respond when i can

<ekaitz>thanks! i may write a little bit about how to do this in the guile docs or something

<wingo>excellent

<ekaitz>:)

<wingo>working on this level is a bit gnarly :)

<wingo>the env vars at least are documented https://www.gnu.org/software/guile/manual/html_node/Environment-Variables.html#index-GUILE_005fJIT_005fTHRESHOLD

<ekaitz>wingo: i think i got something

<ekaitz>how do you call the data you introduce between the instructions?

<ekaitz>not immediates but from one instruction to the next

<ekaitz>i think it's read in the wrong byte order or something

<ekaitz>=> 0x0000003fbaa65058:»·jr»·t2

<ekaitz> 0x0000003fbaa6505c:»·unimp

<ekaitz> 0x0000003fbaa6505e:»·unimp

<ekaitz> 0x0000003fbaa65060:»·lw»·a5,84(a0)

<ekaitz>those unimp are the values i need to read

<ekaitz>0x3fbaa6505c:»··0x00000000»·0xbaa6497c»·0x0000003f»·0x00000397

<ekaitz>that's what the memory has

<ekaitz>and it loaded this to t2 0xbaa6497c00000000

<ekaitz>i think that should be: 0x3fbaa6497c

<ekaitz>instead

<ekaitz>so imho there might be two problems here: the immediate is not inserted properly

<ekaitz>and it's not properly read

<ekaitz>the ld that loads it is wrong

<ekaitz>yep, exactly

<ekaitz>emit_veneer looks problematic

<ekaitz>no, it's in the patch, that's patching in the wrong place

<ekaitz>ACTION runs but will try to fix it later

<jfred>God I love how hackable guile programs can be. I was just trying to add a table to my haunt blog post using the skribe reader, and realized that the skribe reader didn't expose the HTML table tags. But I peeked at the source and noticed that the skribe reader can take a 'modules' param that lets you load your own. Couple minutes later, and now I've got tables

<jfred>thanks dthompson for making haunt so malleable

<dthompson>jfred: glad you're having a good time :)

<marmalade>rlb daviid tyty

<ekaitz>wingo: it was a packing issue in a veneer

<ekaitz>that uncovered another problem :)

<old>anyway to use parameterize around top-level defines?

<old>e.g: (parameterize ((foo 1)) (define-public fuz 1))

<old>looks like fuz is not been exported like I would've expected

<dpk>there is no splicing-parameterize

<dpk>can you say why you want to do this?

<dpk>what you probably want is (define your-variable (parameterize ((current-whatever something)) ...))

<dpk>but it won’t close over the parameterization …

<old>I define a parameter P and a GOOPS class C in module M1. C use P as a #:init-value. I want to be able in M2 to change P around a set of definitions

<old>I guess I could just write a syntax and use it to set the #:init-value to something else implicitely

<old>in M2

<dpk>you should still be able to parameterize around something more specific than the definition itself

<dpk>i’m not familiar enough with GOOPS to say

IRC channel logs

2024-11-07.log