IRC channel logs
2024-12-21.log
back to list of logs
<lechner>mwette / Okay, I'm just trying to load a shared library. Can I access the libraries under LIBRARY_PATH via some option to foreign-library and friends? <lechner>Where are the Guile-specific extensions located? <lechner>This conversation reminds me of Libtools -module <lechner>mwette / wouldn't that also make Nyacc's FFI helper easier to use? <mwette>the guile built-in ones are `(assq-ref %guile-build-info 'extensiondir)' <mwette>I think there is also a site specific one. <mwette>ffi-helper would only use stuff that's a shared library (i.e., libfoo.so) for common C library <efraim>rlb: I don't think JIT is enabled for powerpc but I've never checked. I checked e2ed33ef0445c867fe56c247054aa67e834861f2 (3.0.9+11commits) after it failed a bisect on i686 and I got a segfault building language/tree-il.go as part of stage0 <dsmith>efraim, Guile git was breaking in the STAGE0 build for about 20 commits or so from mid Nov to mid Dec. <dsmith>Starts working againg at 47807c9b118c190fbb6487b2b028170861d7cf5b <efraim>thanks, I'll write that down as a known good commit <dsmith>Starts failing at 7379049d3093fe2076f110c35d482a2aa96d6496 <dsmith>That 4780 commit just reverts the 7379 commit <efraim>unfortunately that's where I'm starting from, 3.0.9 <rlb>I'm pretty sure we had to have that for powerpc. And I (clearly) should have gotten that upstream already, if that's the problem... <rlb>Was powerpc working for you previously? <efraim>I have 3.0.9 working. I hadn't tested since working on the previous failure between 3.0.2 and 3.0.7(?) since I moved my home office and ran out of outlets <rlb>Oh, then if you haven't tested since 3.0.7-ish, I bet you need that patch. <efraim>I have 3.0.9 since it's working in guix, I figured I'd check later commits too <rlb>Hmm, so either guix has that patch too or doesn't need it, or we don't need it anymore... <efraim>I have the 0008 patch but only for stage0/Makefile.in <rlb>I'll try current main again on perotto with and without that patch... <efraim>hopefully its faster than my g4, I measured a successful build in days <rlb>Oh, yeah, I think it's a good bit faster than that, though it is in a qemu envt on whatever it is. <efraim>I'd assume that's basically native, being powerpc64 <rlb>If we know a latest known working commit on main, we can of course (and perhaps you already are) automate some probing via rebase exec, or bisect, or something. Though if there are any changes there requiring a re-bootstrap, that's going to make it "ugly" if we don't know where they are. <dsmith>efraim, I feel your pain. Back when the bot was on a mipsel box, rebuilding Guile (and Bobot++) took at least a whole weekend. <rlb>still does for some archs :) <rlb>ok, main build with that patch is running -- should know if it's going to fail in the same place in an hour or two iirc.