IRC channel logs
2026-04-30.log
back to list of logs
<rlb>dsmith: see also mmdebstrap. <rlb>One of the main tricks of debvm is that it can launch a vm without (partitioned) a disk image --- behind the scenes it reaches into the file, which is just an ext4 filesystem, extracts the current kernel and runs qemu/kvm with that. <rlb>Which also means that you can trivially ftruncate FILE and then resize2fs FILE to enlarge the filesystem if you need to :) <rlb>I also figured out how to get proper qemu virtio directory sharing with the host (via the newer support) *and* accelerated 3d graphics whenever that's useful... So shout if that might be useful and I'll dig up the args. <rlb>(oh, and raw device sharing with the host, for say messing with an untrusted usb drive or other external device) <rlb>...one notable downside (until/unless we fix all the build deps) to our use of include-from-path in module/ (say (include-from-path "sxml/upstream/input-parse.scm")) is mysterious failures after a fix because you forgot to "touch module/sxml/ssax/input-parse.scm" too. Not that I would ever forget that. <rlb>ACTION might make a patch to add deps for all of those... <cow_2001>where's ekaitz? this is beautiful :< "Comment <- ’(*’ (Comment / !’*)’ .)* ’*)’" <apteryx>is there something in guile-ssh already that mimics the algorithm that ssh itself does to try keys; e.g. getting candidates from the system? <ekaitz>okay so forgejo actions and woodpecker are down, I cannot test the CI in lightening <ekaitz>in the last 24h the uptime of woodpecker is 0% <regtur>a branch where I have sublimated all of my PEG enhancements <ekaitz>i read a few of the changes earlier, and they were really cool <regtur>maybe cow_2001 and jcowan are interested <ekaitz>i remember when i read some parts the first time I got interested on the tracer and had a few ideas but i don't remember if I took that a little further <regtur>without tracing PEG is nearly unworkable for any serious grammar <ekaitz>regtur: ofc the first commit we should apply anyway... maybe in a separate PR? <ekaitz>i that could work, at least those weird trailing spaces would be removed <ekaitz>about the rest of the thing, how did you choose the @@ sign? did you take a look to what c peg implementation does? <ekaitz>ACTION has to go but will check the PR later <cow_2001>jcowan: that if then else is bonkers. :| <cow_2001>ekaitz: got to the theory part and it's getting steep <ekaitz>cow_2001: you can skeep that theory part, just the beginning is interesting <ekaitz>the rest of it, if you are not a parser enthusiast doesn't deserve the effort <cow_2001>regtur: oh yeah! that's the other complaint i have! no errors but #f. i am not smart enough to actually fix that, though <ekaitz>you could apply that pr locally and have some fun :) <jcowan>cow_2001: yes, no one would actually write it. But the alternative is to have more reserved words than any standard should be allowed to have. <jcowan>There are about half a thousand reserved words in the latest Cobol standard, and IBM implementations add even more. <mwette>AI says 92. But need to add int..._t etc. <mwette>Cobol is 23 on tiobe index, beating out Ruby, Julia, LISP, Haskell. <mwette>I didn't see Common Lisp or Scheme, so I assume the family of `paren' languages. <rlb>ACTION checks... yes? <ekaitz>i would do it myself but I don't have commit access lol <rlb>I can't right this second, but I'll plan to look at it later when I can. Do you know what the relationship with "adding extensions editing woth emacs" is (in the commit message)? <ekaitz>i would just remove the text from the commit <ekaitz>trailing spaces are bad, we don't need any reason to remove them <janneke>rlb: yeah, "with emacs" is kinda silly, although the guile and/or guix manual talks a bout "the perfect setup", and using `(add-hook 'before-save-hook 'delete-trailing-whitespace nil t)' is a common thing to do? <rlb>Oh, OK, so that's what they meant --- I just didn't understand what the context was. <janneke>but i agree with ekaitz that mentioning emacs is overly specific, and unnecessarily so <rlb>And, dunno. I wouldn't turn that on by default because different projects have different policies with respect to otherwise unneeded whitespace changes (can be a mess wrt merge conflicts, until unless they're all cleaned up, and enforced --- for active projects). <janneke>ACTION worked with regtur on these commits, although regtur is really the author <ekaitz>still need to review the PEG changes in detail but they look very cool <rlb>But I think it's fine here, unless we have some policy like that. <rlb>(And it won't cause a mess wrt utf8, which is the only giant thing, other than whippet, that I know of that might be "out of tree" wrt rebasing.) <janneke>ekaitz: re "@@" (did regtur answer you?) <janneke>anyway, regtur initially chose `#', but then "someone" got their patch in early and occupied that character for comments <janneke>so then regtur chose `@' (the second @ is a texinfo escape) <janneke>fall-back parsing is a unique feature, there's nothing to be seen about that in the C (or other) PEG variants ;) <mwette>Removing trailing whitespace makes the source more compatible with editors used by contributors. <ekaitz>the C PEG variant has a lot of extra magic, like expression execution and stuff, some of that could be included <janneke>fall-back parsing in PEG allowed us to implement syntax highlighting and auto-completion for dezyne <regtur>I guess I am a bit late to the party janneke <regtur>ekaitz: I would not be in favor of expression execution, which I assume are semantic actions, in the context of backtracking parser <ekaitz>it's still interesting to take a look to what other people do <regtur>Undoing your semantic action when parsing fails after executing the semantic action makes things very cumbersome <regtur>BTW the `@' is for the expect parsing, as in expect non-terminal at <regtur>the fall-back parsing is built on top of the expect operator <regtur>janneke: claimed fall-back parsing allowed us to implement syntax highlighting and code completion, but forgot to mention that it even works for partially valid input <ekaitz>how really? like having partial matches and re-sync ? <ekaitz>because with syntax hihglighting and completion basically what you have is a program that is mostly wrong all the time <ekaitz>the approach to parsing it is different from what parsing a file should be