sb0 changed the topic of #m-labs to: ARTIQ, Migen, MiSoC, Mixxeo & other M-Labs projects :: fka #milkymist :: Logs http://irclog.whitequark.org/m-labs
key2 has joined #m-labs
acathla` is now known as acathla
acathla has quit [Changing host]
acathla has joined #m-labs
key2 has quit [Read error: Connection reset by peer]
rohitksingh has joined #m-labs
Mon_ has joined #m-labs
Mon_ is now known as Guest46773
Guest46773 has quit [Client Quit]
Mon_1 has joined #m-labs
Mon_1 has quit [Quit: This computer has gone to sleep]
sb0_ has quit [Ping timeout: 264 seconds]
sb0_ has joined #m-labs
Mon_1 has joined #m-labs
Mon_1 has quit [Quit: This computer has gone to sleep]
Mon_1 has joined #m-labs
sb0_ has quit [Ping timeout: 256 seconds]
Mon_1 has quit [Quit: This computer has gone to sleep]
Mon_1 has joined #m-labs
sb0_ has joined #m-labs
rohitksingh has quit [Ping timeout: 250 seconds]
cr1901_modern has quit [Read error: Connection reset by peer]
Mon_1 has quit [Quit: This computer has gone to sleep]
<whitequark>
sb0_: regarding binutils, I think _florent_ uploaded it
<whitequark>
we don't really need to rebuild it so I'm ok with just letting it be
<whitequark>
sb0_: regarding py35_someversion vs py_someversion, I vaguely recall another conda bug that forced me to do this
<whitequark>
I'm pretty sure it did not work with the default py35_whatever
<whitequark>
feel free to adjust it as needed, I have no particular insight into it
<GitHub14>
[artiq] jordens pushed 1 new commit to subprocess-termination: https://git.io/vgoYw
<GitHub14>
artiq/subprocess-termination 6434a9c Robert Jordens: Merge branch 'master' into subprocess-termination...
<whitequark>
sb0_: regarding icu being in the dev channel, this is my mistake that I forgot to fix. the buildbot uploads everything to the dev channel by default.
<whitequark>
I will fix that now
<whitequark>
regarding artiq package on windows, yes, I did try it and I think I could not because no 64-bit binutils were available
<whitequark>
anyway a win-64 binutils package will need to be added.
<whitequark>
I'll just hack up the win-32 version since I've never used autotools on windows before and I sure as hell don't feel the need to figure it out now
<whitequark>
especially given that I don't anticipate ever rebuilding the binutils package
<whitequark>
wow, conda's new web interface is even more obnoxious than the old one.
<whitequark>
anaconda.org's.
<GitHub44>
[artiq] sbourdeauducq pushed 1 new commit to master: https://git.io/vgoOY
<whitequark>
before rjo complains that there's too much state in buildbot, I'll say: there is deliberately that much state, downloading all the gigabytes of shit on which artiq depends takes a lot of time
<whitequark>
it would be very easy to clean all miniconda caches every time or even reinstall it from scratch
<whitequark>
let me see if this can be helped by cleaning just the local packages
<whitequark>
ok, cleaning the built packages works
<whitequark>
what
<whitequark>
what the fuck is your problem conda
<whitequark>
what the fuck does this even MEAN? Error: ERROR: placeholder '/var/lib/buildbot/slaves/debian-stretch-amd64-3/miniconda/envs/_build_placehold_' too short in: qt5-5.5.1-0
<whitequark>
hahahaha wow
<whitequark>
what it tries to do is to patch paths embedded in binary files
<whitequark>
and this "placeholder too short" error means that it is unable to do so because there's not enough space
<whitequark>
what kind of idiotic idea that even is in the first place
cr1901_modern has joined #m-labs
Guest38665 has quit [Quit: This computer has gone to sleep]
<whitequark>
I actually don't know how to fix that, it doesn't seem to be adjustable.
<whitequark>
sb0: rjo: unless you can find some better solution, the only one I see is to completely redeploy buildbot from scratch in a less nested directory
rohitksingh has quit [Ping timeout: 272 seconds]
rohitksingh has joined #m-labs
<whitequark>
actually, this mmmmmight also be responsible for the windows breakage, though i'm not sure yet
<whitequark>
sb0: ok, I hacked up a binutils-or1k-linux package for win-64
<whitequark>
given that it's built using cygwin, there is no difference between them anyway
<whitequark>
now the main problem is that the windows conda doesn't correctly set the PATH for binutils to be found...
<whitequark>
so many conda problems
<whitequark>
ok, so now there are only these two conda problems.
<whitequark>
mhm, needs to be installed into Library/ subdirectory on Windows apparently...
Guest81932 has quit [Quit: This computer has gone to sleep]
<sb0>
anything that can go wrong will. yesterday i debugged a artiq installation that turned out to be running on a 32-bit machine, and it was kinda half-working as it found some old 32-bit packages that were laying around in our anaconda page
<sb0>
the fact that it installed at all on that machine gave the impression that 1) 32-bit linux is supported 2) many things in artiq are broken
<sb0>
whitequark, oh do i have to use conda's "activate" for qt to work?
<whitequark>
duh
<whitequark>
I would kill those 32-bit pkgs but it's extremely annoying in anaconda.org
<whitequark>
you have to manually go through pages and pages of crap and carefully select it
<sb0>
I have done it already
<sb0>
yes
<whitequark>
ok
<sb0>
how do you use activate on windows?
<whitequark>
activate artiq
<whitequark>
batch files always run as if they used 'source'
<sb0>
in what directory? the problem right now is it doesn't find the activate script
<whitequark>
....\miniconda\Scripts
<whitequark>
no bin on windows
<whitequark>
keep in mind I just deleted the qt5/pyqt5 packages, so any fresh installs will not work
<whitequark>
well, not like they worked before
<whitequark>
anyway you'll have to wait until it builds.
<sb0>
looking at the DLL problem rn
<whitequark>
it should just be the 'activate'
<sb0>
the old package should still be there
<sb0>
got activate to work
<sb0>
vnc is a pita on lousy train wifi
<whitequark>
not vnc, xen's broken vnc
<whitequark>
but yes
<whitequark>
didn't you have ssh?
<sb0>
why didn't you use Xvnc and run QEMU in a X11 window?
<whitequark>
because qemu is mortifyingly slow, and the whole point of this is to compile
<sb0>
okay it seems to import now. so the solution to the DLL problem is to use activate.
<sb0>
yes, that worked
<whitequark>
bb-m-labs: force build --props=package=qt conda-lin64
<bb-m-labs>
build #54 forced
<bb-m-labs>
I'll give a shout when the build finishes