IRC channel logs
2026-04-22.log
back to list of logs
<dsmith>old, Yes, several tests are now skipped on fbsd. Because they don't implement what the test is looking for. <jcowan>rlb: No, but it should be possible to trivially adapt the SRFI 143 code <dsmith>old, The main thing was getting configure to use the correct spawn implementation <rlb>Oh, right, and yeah, that's what I was leaning toward. <rlb>It'd either be internal only, or perhaps, an guile extension. <rlb>Anyone know the story behind ice-9/test.scm and r4rs.test? And do we use the former anymore anywhere? <spk121>rlb: r4rs still gets run in make check. test.scm appears unused, but, you can still run it. But you need to copy it into your current directory. test.scm goes way back. mid-1990s <mooseball>does geiser have a way to do backreferences without knowing the $ number? like when my return value is huge, i have to manually scroll back up to find out what its $ number is. with SLIME i would just use * or ** or ***, working from most recent backwards <old>mooseball: Something like _ in Python? <old>Sorry I don't know SLIME so .. <dthompson>geiser doesn't do anything special here. the $ variables are provided by (ice-9 history) and afaik there are no back references. <old>my tricks is this. Just enter a number or string in the REPL, you will get $N = X <old>but I think we should probably add back referneces if possible that would be more nice <mooseball>dthompson: thanks for explaining, that's why i couldn't find anything on it in the manual <mooseball>it'd be nice, as 99% of the time i just want to refer to the last value, which never has the same number <old>I also tend to have some weird warning about undefined variable when using a $X value when I am in a nested REPL (exception) <old>I'll be happy to investigate see if I can add it for next release <mooseball>old: i haven't seen the undefined variable error, but i'm a pretty basic user of geiser/guile, maybe i'm not making fancy enough exceptions. thanks for the help <jcowan>Backrefereces crap on regex performance: they should be avoided. <old>can you expand on that? <jcowan>They make it impossible to use Thompson DFAs, which are much more performant than backtracking NFAs. <jcowan>The SRFI 115 library is an example of this approach. <old>I am not sure to follow here. How is defining a variable like $$ for the last value returned a problem for regexp? <identity>either jcowan misunderstood the kind of backreferences you were talking about, or it was an unsolicited opinion; either way, backreferences should be avoided in regular expressions :3 <jcowan>All my opinions are unsolicited. <jcowan>However, backreferences like \1, \2, ... in the regex are the problem. If you mean substring matches, they impose a cost as well, but not as bad. <identity>jcowan: they were talking about backreferences in the REPL; when Guile prints «$N = value» it actually does (define $N value) or whatever <old>regarding the REPL output, I was also thinking to add a knob for limiting the time it run and size <old>really painful when the REPL is printing a 2Mb ELF file bytevector <identity>«this value is large as hell, do you really want to print it?» <jcowan>CL has two knobs, one for depth, one for width <jcowan>(parameters / dynamic variables) <old>identity: it's kind of difficult to determine how large a value will be (minus if you have the string already) <old>but I guess for bytevector that would be doable <jcowan>they are spelled *print-level* and *print-length* <jcowan>and they output # and ... respectively to prevent rereading <jcowan>In Scheme we'd need #... when *print-length* is in effect <jcowan>All SRFI texts and code are MIT unless otherwise mentioned <jcowan>If you want, you can add a per-file license <old>rlb: althought we have not agree on that yet, I would tend to just go with: <old>;;; SPDX-License-Identifier: MIT <rlb>Right --- I think we (guile) require one, so I'd need to add one. I just wanted to remember how we should/could handle that wrt guile policies, i.e. do I just add one. <old>;;; SPDX-FileCopyrightText: Author <old>But I think we currently require full license preamble .. <rlb>OK, well, I'd definitely prefer SPDX I think (very helpful downstream), but I mostly meant wrt being allowed to assume the implict copyright, and/or determining the authors, etc. <rlb>i.e. do I git blame the entire file, or...? <rlb>i.e. when the file has one, of course we can "just take their word for it", but when I have to make one, wanted to be sure I did whatever we prefereed. <old>the file is understand version control? <rlb>(according to git blame "Arthur A. Gleckler" write most of it, but that's a commit that says "Apply change from John Cowan" which wholesale adds the file ;) ) <rlb>i.e. the reason I asked <old>it shis copy/paste from actual srfi-152 or there are bits that are unique to chibi? <rlb>Now I think the right answer for that file is probably mostly jcowan, but wanted to know what the process should be if there is one. <rlb>That *is* the official srfi 152 repo <jcowan>Arthur is the editor of all recent SRFIs <rlb>Basically just wanted to know what we do with a srfi file that has no per-file copyright, i.e. do I do git archaeology, and if so, how much (just log %aN, or deeper checks like I just did, or...) <jcowan>so commits from him are usually not substantive <jcowan>The SRFI document sometimes lists the implementor near the top, but if not, it's the same as the author. <rlb>Basically means the git history is only potentially useful wrt determining a proper copyirght header/years/etc. <rlb>(i.e. there the "year" isn't clear because all we have is the date Arthur committed the initial version.) <rlb>If we're really trying to be careful about the file copyright header dates... <jcowan>Not like it matters what the date is: copyrights expire at life+70 <rlb>old: I can also just try to do a reasonable job and make what I did clear in the initial commit message I suppose. <jcowan>AFAIK no SRFIs are copyrightedby corporations <jcowan>The tests use (chibi test) if available, or failing that SRFI 64 <jcowan>chicken-test.scm uses the chicken (test) library <rlb>I'm primarily just concerned with trying to make sure I respect guile's policy, so I can "proceed". <rlb>jcowan: right, I diffed the three test files, and they're nearly identical, so I just picked chibi. <old>rlb: I think so yeah <rlb>OK, thanks (and had already gotten started with that assumption ;) ) <rlb>I think it may just be that file anyway, so less critical --- i.e. think the others we'll want do have per-file declarations. <jcowan>Because SRFI 152 is fourth-generation, the code is originally Shivers's, modified by Shinn, modified by Clinger, modified by me. <rlb>right, the other files I think mention some of that I think <rlb>And thanks for the confirmation wrt the tests. Any particular year(s) you'd prefer in the test file copyright? <old>rlb: I'm pushing with-modules in the next 5 minutes <old>you will have to recompile psyntax but otherwise it should be okay <old>I just need to remember how to update psyntax-pp now .. <rlb>doh (wrt remembering), and sounds good <rlb>sneek: later tell spk121 thanks --- mostly (aside from curiosity) just wondered if ice-9/test.scm wasn't actually used/tested by anything, whether we wanted to keep it (in addition to r4rs.test), but not a big deal either way. <old>I've pushed the `with-modules' bits on main if anybody wants to try playing with it <old>usage examples can be found in: test-suite/tests/syntax.test <old>try to break it please :) <rlb>I did at least try a local build/make-check, and that was just fine. <rlb>I'll also try it with utf8 soonish. <rlb>(though can't imagine why that would matter)