<rjo>
but i think there are several usages of list_targets() that need to be made slice-accurate
ylamarre has quit [Quit: ylamarre]
evilspirit has joined #m-labs
<sb0>
whitequark, is the source somewhere?
<sb0>
those figures look surprising and i'd like to do a quick check ...
<whitequark>
asked
<sb0>
rjo, hmm there are also slices of slices (though i think verilog will break for those too)
<sb0>
rjo, and btw i never thought about having a signal appear in both comb and sync - even if it's different slices of it. not sure if other things break if you do that...
<sb0>
(appear as target)
rohitksingh has joined #m-labs
ylamarre has joined #m-labs
ylamarre has quit [Ping timeout: 250 seconds]
rohitksingh has quit [Ping timeout: 246 seconds]
<acathla>
from migen.fhdl.std import * does not work anymore...
<acathla>
ok, from migen import * is ok
<cr1901_modern>
acathla: There was an API change a few months back
<acathla>
cr1901_modern, is there any doc about it?
<acathla>
except the README?
<cr1901_modern>
Yes, but it's incomplete. Lemme find t
<acathla>
I'm starting again the tutorial and adaptation to the avnet board
<acathla>
cr1901_modern, thank you.
<cr1901_modern>
np... I have code that I need to rewr- err, update for the new imports myself :P
ylamarre has joined #m-labs
rohitksingh has joined #m-labs
rohitksingh has quit [Ping timeout: 260 seconds]
ylamarre has quit [Ping timeout: 240 seconds]
rohitksingh has joined #m-labs
ylamarre has joined #m-labs
<GitHub113>
[artiq] jordens pushed 1 new commit to master: http://git.io/vua0z
<GitHub113>
artiq/master 358ad2e Robert Jordens: artiq_flash: drop redundant instruction, tweak doc
<GitHub162>
[artiq] jordens pushed 1 new commit to master: http://git.io/vua0j
<GitHub162>
artiq/master 73b4aff Robert Jordens: README: drop old travis badge
<rjo>
sb0: if it is really that uncommon (to have signals split between comb/sync), maybe migen (at least in the simulator) should enforce that. also seems needed to optimize the execution of fragments (modified subset of the signal dependency graph).
<rjo>
whitequark: can you give me push access to the buildbot or auto-trigger the builds?
<rjo>
whitequark: also how is the hostname for that machine coming along?
evilspirit has quit [Ping timeout: 260 seconds]
<whitequark>
sb0: ^
<rjo>
whitequark: let's do this here.
<rjo>
whitequark: ic.
<rjo>
whitequark: anyway. once there is a DNS entry, we should not auto-upload the conda packages. that should still need forcing.
<whitequark>
why so? the dev channel exists specifically for auto-uploading.
<whitequark>
they shouldn't appear on the main channel though.
<rjo>
fine. then the documentation should differentiate between the two.
<whitequark>
sure.
<rjo>
(it doesn't)
<whitequark>
I think that's because artiq/migen/misoc haven't really had any useful releases yet.
<whitequark>
they should, but I'm afraid it's not my call as to when make one.
<rjo>
and imho recommending conda envs with date metadata in the env name is asking for trouble. just a conda install artiq the next day will make that meaningless.
<rjo>
and iirc the main channel (by itself) was broken last i checked. but i forgot what was missing.
<whitequark>
yes, it does not have any artiq releases in it.
<rjo>
"when it's ready" ...
<whitequark>
as for conda envs with date metadata in the name, iirc the docs tell to not do anything with the environment after it's created.
<whitequark>
but it's probably true that this advice will be disregarded and it might not be the best way to manage environments
<whitequark>
if I remember correctly, I wrote that after someone asked for a way to make sure that upgrading artiq always leaves a way to go back to a working version.
<whitequark>
since conda does not include version lockfiles, like any other decent package manager in existnece (bundler, npm, cargo, ...), it does not seem to me that there is a way to do this apart from what I currently suggest in the docs
<rjo>
then i would do artiq-prod and artiq-(dev|stage|test) and act accordingly.
<whitequark>
ideally, conda would remember the versions of packages used to successfully run tests and put them in some file
<rjo>
the trigger for a rollback in the lab is likely not a failing unittest
<whitequark>
oh, yeah, I phrased that poorly. I meant "the latest versions of packages at build time"
<rjo>
doesn't "freeze" do that?
<rjo>
or whatever it's called?
<whitequark>
what is "freeze"?
<rjo>
conda list
<whitequark>
ahh, conda list --export. I see.
<rjo>
and for the buildbot the trigger-graph is migen->misoc->artiq->artiq-* ?