IRC channel logs


back to list of logs

<lfam>Testing QEMU, I saw the SSH host key generated in the VM image super-early in the boot process
<sturm>Is there a way to view the build output for a package on Hydra, just to confirm that it's seeing the same errors as I am?
<sturm>Wonderful, it has a handy dandy web interface:
<lfam>It's the first time anyone called it "handy dandy" :)
<taohansen>ldd is part of which package?
<CharlieBrown> is sexy too.
<efraim>taohansen: iirc ldd is part of gcc-toolchain
<efraim>ld is in gcc-toolchain, actually I don't remember for ldd
<luther9>hey, i'm trying to run g++, but i'm getting weird errors. is there a guix equivalent of debian's built-essential?
<efraim>There's gcc-toolchain and there's 'guix environment foo' where it will create an environment with all the dependencies and build dependencies for foo
***pksadiq_ is now known as pksadiq
<civodul>Hello Guix!
<civodul>rekado: the other day, did you install GuixSD from an ISO image?
<civodul>as in "disk-image -t iso9660"
<rekado>I have not tried an ISO, only the raw disk image.
<civodul>i just found a (small) regression introduced with overlayfs
<civodul>i'm running the system tests and i hope to start updating NEWS today
<brendyn>hmm i was pretty sure a cc'd the mailing list with my responses to the MIME data thread but they didn't appear
<rekado>R 3.4.3 is scheduled to be released tomorrow. I’m going to update it and all of our CRAN/bioconductor packages.
***pksadiq_ is now known as pksadiq
<efraim>didn't have a chance to look at t1lib yesterday, got hit by a car (everyone's ok) and just now finished with the police
<efraim>So I probably can work on it today
***jonsger1 is now known as jonsger
<brendyn>ok several hours later my emails come through. i thought the list updated at at least every half hour
<civodul>rekado: i found a hack to get rid of the GUILE_WARN_DEPRECATED messages during 'guix system build' \\o/
<civodul>mb[m]1: thanks for the libxcursor/libxfonts fixes!
<civodul>and lfam for the QEMU ones
<dustyweb>hey civodul
<dustyweb>civodul: not sure if you saw my reply to yours about the nginx / certbot stuff
<dustyweb>sorry I didn't reply sooner :(
<dustyweb>I'm excited certbot is in but I suspect we need to deal with this before the next release at least
<dustyweb>I may have some time to work on it next week
<efraim>i see the 'no grub tests on arm' patch was pushed, it really should be 'intel-only' for the tests on vanilla grub
<nee`>mariadb should probably be updated. Guix currently has 10.1.26 and the latest stable versions are 10.1.29 and 10.2.11
<wingo>civodul: where is that python.scm file you were using for testing?
<wingo>ACTION has a little fix
<wingo>civodul: how long does compiling python.scm take for you? for me is 6 seconds
<efraim>civodul: looking at t1lib, it looks like a couple of programs embedded it, and have since dropped it, except for evince, so I think it's safe to remove it
<wingo>ok, merged stable-2.2 to master, and made a small speed tweak to stable-2.2 also
<CharlieBrown>Would it be a bad idea to use Guix GNOME on a foreign distro?
<wingo>hehe, that python.scm number was only for compiling to cps, not bytecode :P
<wingo>CharlieBrown: certainly ok to use gnome apps; dunno about the env as a whole though
<CharlieBrown>I know GNOME can be used on GuixSD, but not sure about how hard it would be to get it to work on a foreign distro.
<wingo>civodul: hack to slot allocation seems to reduce python.scm compilation time from 52s to 38s
<wingo>not great but better
<civodul>wingo: neat!
<civodul>my feeling is that we'd need a different implementation of graphs, at least one with both edges and back-edges to avoid the repeated revert-graph calls
<civodul>but i haven't been able to experiment with that yet
<wingo>possibly, tbh tho i think improvements to the algorithm especially for -O0 would be almost just as useful
<civodul>re python.scm, i did some of the benchmarking on just the cps->bytecode part
<civodul>without optimizations
<wingo>i.e. it should be possible to do a simple once-through slot allocation instead of so much flow analysis
<wingo>for -O0
<wingo>might not get ideal slot allocation but it would be fine
<wingo>anyway our slot allocator is not ideal anyway :P
<civodul>the top-level of python.scm is typically something where slot allocation doesn't matter that much, no?
<wingo>civodul: indeed!
<wingo>i think we could avoid some more time by short-cutting the lazy variables thing
<wingo>which is just a form of pre-coloring i guess
<wingo>that takes 6 seconds for the big function
<wingo>a little more than computing live variables :P
<civodul>good that you find new optimization opportunities :-)
<wingo>if i just return empty-intset from compute-lazy-variables i get down to 30s
<civodul>probably a smaller heap as well?
<wingo>could be
<wingo>where is that "time" in guix?
<wingo>the one that lists max heap size
<civodul>"guix package -i time"
<civodul>but i found that displaying the heap size reported by libgc is sometimes more useful
<wingo>ACTION nod
<wingo>reports 2G max resident, if i read that right
<wingo>can that be possible??
<wingo>top shows a max virtual of around 650MB and max resident of around 500MB
<wingo>but that's top :P
<wingo>dunno how to compare that to what you have seen tho
<wingo>that's the profile-guilec script but compiling to bytecode instead of cps
<civodul>yeah i don't know why time(1) reports values this different from what top suggests
<CharlieBrown>Uh-oh. Trisquel's GNOME did not work at all. Any chance Guix's could be used instead?
<civodul>CharlieBrown: not sure if you have something specific in mind, but GNOME on GuixSD does work
<civodul>ACTION wonders why curl depends on openldap
<civodul>maybe for the kitchen sink user directory?