IRC channel logs
2025-11-26.log
back to list of logs
<old>AFAIK, golang revert from that <old>bootstrapping what exactly? <old>glibc do so many things it's basically ABI on Linux <sneek>tohoyn, daviid says: I 'll patch scm->gi-array and gi-array->scm so they work fine with array of GVaraint myself, tx - I will kepp GBDus struct semi-opaque though, so you can write your own api to build those you need (and any other user) <janneke>so, how do quasipatterns work in match? i find the docs to describe the syntax, but not the semantics? <janneke>scheme@(guile-user)> ,m (ice-9 match) <janneke>scheme@(ice-9 match)> (match '(key value) (('key v) v)) <janneke>scheme@(ice-9 match)> (define dynamic-key 'key) <janneke>scheme@(ice-9 match)> (match '(key value) ((`,dynamic-key v) v)) <janneke>scheme@(ice-9 match)> (match '(foo value) ((`,dynamic-key v) v)) <janneke>$3 = value ;; Argh, no; should not match! <civodul>janneke: i think it doesn’t work the way you think <civodul>that is, (match '(foo value) (`(foo ,v) v)) does what you expect <civodul>to match on a variable, AFAIK you have to resort to a predicate <civodul>like ((? (lambda (x) (eq? x dynamic-key))) value) <ieure>I was looking at this the other day, and I agree with civodul, there doesn't appear to be an elegant way to match a specific value like that. <ieure>"elegant" meaning "syntax less verbose than a lambda predicate" <mwette>match redefines the meaning of quasiquote and unquote <ekaitz>we need to improve the documentation for that <sneek>I've been a process for one month and 17 days <sneek>This system has been up 6 weeks, 5 days, 18 hours, 56 minutes <rlb>civodul: is "make -j5 distcheck" supposed to work from a clean configured tree? Here it crashes, I think (given the interleaved output) trying to build auto-conf-macros.texi with "/usr/bin/guild: 3: exec: /home/rlb/src/guile/up/main/libguile/guile: not found <rlb>I wondered if this might just be an existing issue, i.e. no one ever tries a parallel distcheck from scratch. <rlb>At least it's easy to reproduce... <rlb>Even a serial "make distcheck" after a successful "make -j5 check" fails like this with main: <rlb>echo 3.0.10.284-35288 > ../../.version-t && mv ../../.version-t ../../.version <rlb>/bin/bash: line 1: ../../.version-t: Permission denied <rlb>Somehow the entire guile-3.... tree that I think distcheck was trying to build was read-only. I'm guessing there's some -j related race in a "from scratch" parallel distcheck. I'll try again with *everything* serialized. <dsmith>rlb, Could it have been trying to write to a system dir, Like /use/local/... ? <rlb>No, that was from within SRCDIR/guile-VERSION-GITINFO/build or something, and guile-VERSION... was read only. <rlb>I'm guessing that some step makes it read only once it's "ready", but some other step that should have gone first is running in parallel. <rlb>That dir might also have been left over from my initial from scratch "make -j5 distcheck" run -- either way, suspect there's a -j related bug. <rlb>If anyone has time, might be nice to have a double-check, though of course if it's a race, then might or might not be easily reproducible "elsewhere". <rlb>I also suspect that this might not be new, and so might not need to hold up the release. <dsmith>yep! In maint.mk in release-prep: is `chmod a-w $(rel-files);` <rlb>Not sure whether it's changed, but for "forever" perl's been one of the few (only) "languages" you're allowed to rely on being available in debian package maintscripts (prerm/postrm/preinst/postinst/etc.). i.e. unless it's changed and I didn't notice, you can't use python etc. Typically just use perl, (ba)sh, coreutils... <ieure>Yeah, perl is in Debian's base package set, I think mostly because a bunch of the early Debian tooling was written in it. <ieure>I think some still is, but maybe less than used to be the case. <rlb>...I'm probably going to be refreshing my perl in a while because the debian emacs infrastructure needs an overhaul. Things I didn't get right the first time (a good while back). <rlb>Yeah, "make distcheck", run after a successful build just crashes on that .version-t write every time. <rlb>I wonder if this is a SRCDIR/BUILDIR issue or something similar. <rlb>Also strange, right after that failure, and "rgrep -F .version-t" in our source tree finds nothing. <dsmith>That chmod a-w was last touched in 2012. I wonder how long it's been since a make distcheck <rlb>Dunno -- if we use that to produce the release archives, then "fairly recently"? <rlb>ACTION wonders if build-aux/git-version-gen really should/needs-to have the unsafe time-stamp LocalVariables... <rlb>Yeah, looks like it's probably a src vs build dir issue -- but what's trying to write that file... <rlb>OK, I think it's the "$(top_srcdir)/.version: Makefile <rlb>" rule at the end of ./Makefile.am. <rlb>Perhaps it needs $(distdir)(s). <dsmith>Why the echo > file-t && mv file-t file dance ? <rlb>I'm testing this now: <rlb>cd $(distdir) && echo $(VERSION) > $@-t && mv $@-t $@ <rlb>Though I don't know if that's actually the right thing. <dsmith>$@ is $(top_srcdir)/.version at that time, no? <rlb>Oh, wait, that won't be if distdir is the actual ./guile-VERSION dir because it's *supposed* to be read only. <rlb>I wonder if it was already broken or I broke it in 6bd70136d96e73542e6725bc490199e17f56ee92 <rlb>If anyone else here is better with auto*, by all means, take a look wrt what we ought to be doing there, otherwise I'll try to delve later.