IRC channel logs
2025-03-11.log
back to list of logs
<dsmith>lechner, So your pam module has some kind of failure when called from a Guile script. What are the condidions where it works? From a C progam? <dsmith>I'm wondering about how libguile is being initialzied. <dsmith>ERROR: Wrong type to apply: "quote" <dsmith>Does that mean "quote" as a string? As in ("quote" ...) ? <dsmith>scheme@(guile-user)> ("quote" 123) <dsmith>ice-9/boot-9.scm:1676:22: In procedure raise-exception: <lechner>dsmith / it works from C (mingetty) and Rust (greetd) although the latter produces an "madvise: failed" message <lechner>dsmith / i think that's unfortunately transitory <lechner>it's a real stress test, i think. it goes from Guile->dlopen->libpam->dlopen->custom C module->libguile->Guile->dlopen->ELF modules <lechner>i am pretty confident, however, that the error occurs in the lower Guile level <dsmith>So something that really help be get to the bottom of a arm jit issue some years ago was disabling "address space randomization" or *something* like that, <dsmith>Unil I disabled that, every run was different. <dsmith>Can't remember the exact thing right now.. <rlb>dsmith: (just guessing) looks like it could be setarch -R? <rlb>ACTION has never tried that <dsmith>rlb, Thanks! That does look suspiciously like what I'm remembering. <dsmith>lechner, Try running your program with that and see if you can get more consistent results <dsmith>Looks like you can also set it inside gdb for the prog you are debugging. <dsmith>set disable-randomizaiton {on,off} <mwette>janus: Did you put a `(use-modules (srfi srfi-9))' in the file before the `(lalr-parser' part? (I don't use (system base lalr) fwiw.) <dsmith>lechner, How are you init-ing libguile in the pam stuff? <lechner>dsmith / was your advice to disable address space layout randomization in the kernel? <dsmith>Just in the failing Guile process. <lechner>a single C module replaces all others in the libpam stack but calls all others as needed via a Guile implementation of the stack. modules can also be written in Guile <old>I have a hook on module-defined-hook. However, the module I get has unbound variables for its public interface <old>any idea how I can get a hook when a module is loaded and its variables are bound? <dsmith>lechner, The scheme_with_guile sure seems right. <dsmith>Hmm. I wonder if disabling the JIT improves things. <dsmith>Or if something is stomping on some stack somewhere <old>hmm I manage to fix my problem <old>but now I have another <old>I'm using traps to do coverage with a better granularity then add-next-hook! <old>however, it seems that if I try to instrument a procedure in a file that was loaded with `load-from-path' I only get instruction pointer from ice-9 eval <jean0t>I want to be a contributor to the project <jean0t>what do I need to do outside of registering in the mail list <jean0t>to be honest I am not familiar with shceme, but Ive been studying common lisp recently <dsmith>Response times are usually measured in days... <dsmith>lechner, And remember, absolutely NEVER let a SCM become all-bits-zero. Your life will become very unhappy.. <vbramselaar>Is there any alternative to using autotools for a Guile project? <vbramselaar>Apart from that, I would love to use Guile as embedded language in a game with SDL2/C++ etc. But I'm just too scared of Windows compat :( <vbramselaar>I did see that MSYS2 has it now, so that is interesting. <old>vbramselaar: there is on going effort to make a build-system in Guile. However, I'm not sure who we will make it work on Windows <vbramselaar>mwette: I know about guile-sdl2, the issue mainly is Windows (I know GNU dislikes Windows, rightfully so of course). <dthompson>vbramselaar: for guile it's moreso that very few developers actually use windows, so the windows build is lacking <old>vbramselaar: I've started working on that with two other peoples yes. It's work in progress <old>many ideas for it. like make it portable to distros where guile is not available <dthompson>for those that care about guile and sdl stuff, sdl3 was recently released and I'm working on guile-sdl3. <dthompson>not only is sdl3 way better than sdl2, guile-sdl3 is going to be better than guile-sdl2. :) <dthompson>basically I need to write the manual and clean up a few things and then it will be good enough for a 0.1.0 release. the bindings will be far from complete but all the structure will be there to hopefully make it easy for others to fill in gaps. <vbramselaar>old: that would be nice. I mean Autotools exists but I feel like a build system specifcally for the language is missing. <dthompson>sometimes I think "I could write a build system" but that is pure hubris, ego at its most extreme <old>was it in any good shape at some point actually <old>seems like a half-baked forge from the 2000s <dthompson>I was trying to be generous but you are very correct <old>other freesoftware forge like sourcehut and codeberg are way better in that respect <vbramselaar>why is savannah still used so much by GNU, but feels like that really updated? <old>I'm mostly on sourcehut, but it seems like all the communities are moving toward codeberg <old>might worth looking at it <old>vbramselaar: ask stallman or the FSF idk <vbramselaar>old: im interested with moving from Gitlab to either Sourcehut or Codeberg. <old>sourcehut works okay for now. at least for my use. But I'm still waiting for the beta version .. since 3 years <vbramselaar>but I'm trying to figure out where to keep images (binary files) of my website instead of Git. <vbramselaar>old: yhea I was following that news a bit and they had some setbacks it sounded like recently <old>yeah. unfortunatelly, it seems that forges are poppular not only for hosting projects, but for DDOS <old>and when you don't have the capital of M$, it can be difficult to invest protecting against DDOS <old>I'm saying this, but Github actually has more downtime than sourcehut <lechner>dsmith / thanks! what does it mean to stomp on the stack, please? <lechner>Hi, what is the easiest way to read (and evaluate) all expressions on stdin, please? <dsmith>lechner, overwriting. Think of those Japanese moves from the 60's where a guy in a lizard costume is smashing all the buildings <lechner>okay, i didn't think that still happened to stacks in 2025 <df>oh, it does, it's just that godzilla has learned to be a bit more subtle (guy in a lizard constume, huh!) <dsmith>Well, it *shouldn't*. But garbeled backtraces and weird values makes me suspcious. <df>there is ongoing an arms race between ways to exploit it and ways to protect against the exploits which has now got extremely complex <freakingpenguin>If I ever teach programming dress up as godzilla and smash the student's code, got it. <df>freakingpenguin: I approve of this didactic approach <lechner>Hi, what's a way to signal to 'read' on stdin that there is no more input, please? C-d does not work <dsmith>lechner, When the read system call returns 0 bytes <old>lechner: check for eof <rlb>sneek later tell civodul would it be plausible to mark the 00-repl-server.test inter-protocol test as unresolved for now or XFAIL? <rlb>sneek later tell civodul built main 402e0dfa33f442ad238a0f82a332efa438538840 and ran make check on debian riscv64 (i.e. with the jit enabled), and there were no issues.