IRC channel logs
2025-06-25.log
back to list of logs
<cow_2001>i would say that guile-picture-language's (@@ (pict) pict->svg-sxml) should be public <cow_2001>if ya wanna draw some stuff outside geiser <wingo>i am going to kill ia64 support. anyone with itanium can stay on guile 3. <Arsen>:^) based on experience with gcc, glibc and gentoo, I don't think anyone will <wingo>status update: ok! so i have managed to get guile's heap to be precisely traceable, with the exception of captured continuations. when a continuation is captured, it pins its referents. this should allow for defragmentation / evacuation. still have to pin some objects if their address is taken, as in hashq/hashv <wingo>but the audit now is not about making sure that all allocations are traceable; now it is to make sure that anywhere we need an object to stay in place, that we actually pin it. <wingo>with luck we are a week out from trying guile with a mostly-moving gc <old>wingo: does this impact delimited continuation? What are the impact on fibers in term of performance? <identity>old: with luck we will know in a week i guess <veqq>there's a guile 4 coming?! <dsmith>veqq, Someday. Sure. The gc should def be a major version bump <wingo>old: it should speed up everything, though who knows <wingo>looking forward to the first results <mwette>This collector can move objects? <mwette>I guess I'm only curious for bytevectors, whose addresses may be passed to c libraries. <dthompson>mwette: when the pointer to an object is needed, that object will be pinned <rlb>If we end up with time (I might contribute if I get to it) we might want to scan the bugs -- I think I recall seeing some things there where we'd said we'd wanted to fix them for 3, but missed it, and with 4 we'll have another chance. <bryce_m>Would it be considered a bug that `define-meta-command` is undocumented? (Or is it documented and I can't find it?) <identity>bryce_m: does not look documented on guile 3.0.9 (it is at system/repl/command.scm:192) <bryce_m>(came across it in the geiser source) <ArneBab>wingo: do you need someone to run the r7rs benchmarks and give you a geometric average of the differences?