02:05
ohsix has quit [Ping timeout: 260 seconds]
02:33
<
sb0 >
whitequark, unittests should be self-contained. right now test_dma_playback_time fails if another test that records the trace hasn't been run before
02:34
<
sb0 >
and it just happens to be run before by accident when you run all the tests ...
02:35
<
sb0 >
whitequark, python -m unittest test_rtio.DMATest.test_dma_record_time test_rtio.DMATest.test_dma_playback_time -v crashes the core device
02:35
<
sb0 >
so many bugs jesus fuck
03:03
<
sb0 >
get_handle → dma_retrieve → DmaPlaybackRequest/DmaPlaybackReply ... could those be more consistently named?
03:12
<
GitHub >
artiq/master fe05aed Sebastien Bourdeauducq: firmware: DmaPlayback → DMARetrieve
03:12
<
GitHub >
artiq/master badbd72 Sebastien Bourdeauducq: manual: update example to use DMA handle API
03:12
<
GitHub >
artiq/master e8f7f8e Sebastien Bourdeauducq: DMA: speed up playback...
03:34
klickverbot has quit [Ping timeout: 240 seconds]
03:39
rohitksingh_work has joined #m-labs
03:53
<
whitequark >
sb0: (self-contained) yes, of course, that wasn't done deliberately
04:20
ohsix has joined #m-labs
07:14
mumptai has joined #m-labs
08:42
<
GitHub190 >
rust/artiq-1.18.0 b2cf6df Simonas Kazlauskas: Add intrinsics & target features for rd{rand,seed}
08:42
<
GitHub190 >
rust/artiq-1.18.0 4495f7e bluecereal: Update if-let.md...
08:42
<
GitHub190 >
rust/artiq-1.18.0 ec356bb Alex Burka: rustdoc: fix doctests with non-feature crate attrs
08:43
<
GitHub >
conda-recipes/master f4eb2fb whitequark: rustc: update to 0.18.0.
08:44
<
whitequark >
bb-m-labs: force build --props=package=rustc conda-lin64
08:44
<
bb-m-labs >
build #290 forced
08:44
<
bb-m-labs >
I'll give a shout when the build finishes
09:13
<
whitequark >
sb0: I don't understand, shouldn't the lab CPU be fast?
09:13
<
whitequark >
so far it seems slower than the CPU in my laptop
09:14
<
whitequark >
oh, no one gave the VMs more VCPUs
09:17
mumptai has quit [Remote host closed the connection]
09:19
<
sb0 >
why are we updating rustc again?
09:20
<
sb0 >
if that really needs to be done, please update the "build from source" instructions in the doc
09:22
<
whitequark >
because it's two months old?
09:22
<
whitequark >
and I suspect it will fix one of the bugs, anyway
09:24
<
whitequark >
oh, hm, no, it's not slow, it's about exactly as fast as the one in my laptop
09:28
<
sb0 >
fixing some user-visible bug is a good reason, two months old doesn't really matter
09:29
<
whitequark >
it does
09:29
<
whitequark >
we have a fork of rust and we use lots of unstable features
09:29
<
whitequark >
in both cases it's far easier to fix small amounts of breakage as it rolls in, as opposed to screwing with it for several days straight every half year
09:31
<
sb0 >
but is it really the best thing to do right now? smoltcp is still at 25k/s, doesn't support keepalive, lots of compiler and runtime bugs, artiq tools don't start on windows anymore, etc.
09:31
<
GitHub >
conda-recipes/master eda00a7 whitequark: rust-core-or1k: update to 0.18.0.
09:32
<
whitequark >
we've almost spent longer discussing this than the time I spent touching rustc
09:32
<
whitequark >
bb-m-labs: force build --props=package=rust-core-or1k conda-lin64
09:32
<
bb-m-labs >
build forced [ETA 47m09s]
09:32
<
bb-m-labs >
I'll give a shout when the build finishes
09:35
<
sb0 >
well, make sure the doc is updated and the drtio firmwares still work
09:36
<
whitequark >
yes, I've updated the doc long ago...
09:37
<
whitequark >
I haven't said that I pushed those changes
09:38
<
whitequark >
and I didn't do that because the buildbot build would have immediately failed
09:38
<
whitequark >
I'm now checking drtio_satellite and after that I'll push
09:38
<
sb0 >
well the buildbot github trigger is also broken...
09:38
<
whitequark >
oh right.
09:40
<
GitHub >
artiq/master db49496 whitequark: firmware use Rust 0.18.0.
09:46
bb-m-labs has quit [Quit: buildmaster reconfigured: bot disconnecting]
09:47
bb-m-labs has joined #m-labs
09:48
<
whitequark >
sb0: fixed.
09:48
<
whitequark >
bb-m-labs: force build artiq
09:48
<
bb-m-labs >
I'll give a shout when the build finishes
09:48
<
bb-m-labs >
build #1467 forced
10:14
<
whitequark >
oh,
*great*
10:14
<
whitequark >
leave it to conda to fuck up a perfectly working compiler
10:14
<
whitequark >
it does something to paths inside rustc and rustc doesn't work anymore.
10:14
<
whitequark >
(rustc obviously did work before because the rustc build process involves running a stage1 rustc on stage2 rustc...)
10:15
<
whitequark >
sb0: do you happen to remember what was the incantation needed to let conda-build keep binaries intact?
10:19
<
whitequark >
sb0: ok, so conda-build changed the default for detect_binary_files_with_prefix from false to true in 2.0
10:19
<
whitequark >
this means that a large amount of packages we currently have in repos will probably not work once they're rebuilt
10:19
<
whitequark >
if rustc is any indication, at least
10:20
<
GitHub >
conda-recipes/master f52c329 whitequark: rustc: tell conda to leave binaries alone.
10:20
<
GitHub >
conda-recipes/master 0e8d2c9 whitequark: rustc, rust-core-or1k: bump.
10:21
<
whitequark >
bb-m-labs: force build --props=package=rustc conda-lin64
10:21
<
bb-m-labs >
build #292 forced
10:21
<
bb-m-labs >
I'll give a shout when the build finishes
10:21
<
whitequark >
bb-m-labs: force build --props=package=rust-core-or1k conda-lin64
10:21
<
bb-m-labs >
The build has been queued, I'll give a shout when it starts
10:24
<
sb0 >
whitequark, thanks, what did you change to fix the github hook?
10:24
<
whitequark >
sb0: edited the hmac invocation in /usr/lib/python2.7/dist-packages/buildbot/status/web/hooks/github.py to pass arguments as bytearray()s
10:25
<
whitequark >
(I hope that by the time debian gets a new buildbot, it will also fix that)
10:25
<
whitequark >
probably will happen when squeeze goes stable, anyway.
10:59
<
bb-m-labs >
build forced [ETA 38m04s]
10:59
<
bb-m-labs >
I'll give a shout when the build finishes
11:18
<
whitequark >
bb-m-labs: force build artiq
11:19
<
bb-m-labs >
build #1468 forced
11:19
<
bb-m-labs >
I'll give a shout when the build finishes
11:19
<
whitequark >
bb-m-labs: stop build artiq wrong
11:19
<
bb-m-labs >
build 1468 interrupted
11:20
<
GitHub >
artiq/master 5324a42 whitequark: conda: update rust dependency.
11:22
<
whitequark >
fascinating
11:22
<
whitequark >
the webhook succeeds but it doesn't actually start a build
11:24
<
whitequark >
bb-m-labs: force build artiq
11:24
<
bb-m-labs >
build #1469 forced
11:24
<
bb-m-labs >
I'll give a shout when the build finishes
11:35
bb-m-labs has quit [Quit: buildmaster reconfigured: bot disconnecting]
11:35
<
GitHub99 >
buildbot-config/master 3a35e96 whitequark: Replace our homegrown conda cache cleanup with `conda build purge`.
11:35
bb-m-labs has joined #m-labs
11:39
<
whitequark >
bb-m-labs: force build artiq
11:39
<
bb-m-labs >
build #1470 forced
11:39
<
bb-m-labs >
I'll give a shout when the build finishes
11:49
<
whitequark >
bb-m-labs: force build --props=package=rust-core-or1k conda-lin64
11:49
<
bb-m-labs >
build #294 forced
11:49
<
bb-m-labs >
I'll give a shout when the build finishes
11:53
<
whitequark >
bb-m-labs: force build artiq
11:53
<
bb-m-labs >
build #1471 forced
11:53
<
bb-m-labs >
I'll give a shout when the build finishes
12:25
<
whitequark >
bb-m-labs: force build artiq
12:25
<
bb-m-labs >
build #1472 forced
12:25
<
bb-m-labs >
I'll give a shout when the build finishes
12:36
rohitksingh_work has quit [Read error: Connection reset by peer]
12:53
<
whitequark >
no, seriously, wtf?
12:53
<
whitequark >
I just cleaned all caches...
12:54
<
whitequark >
bb-m-labs: force build --props=package=rust-core-or1k conda-lin64
12:54
<
bb-m-labs >
build forced [ETA 3m57s]
12:54
<
bb-m-labs >
I'll give a shout when the build finishes
12:55
<
GitHub >
artiq/master af20e7e whitequark: conda: update rust dependency.
12:55
<
whitequark >
what a monumental waste of time conda is.
12:56
<
whitequark >
sb0: what's our strategy for dealing with OOM?
12:57
<
whitequark >
rust's collections currently do not allow to detect this (and never will), and they also do not allow to change allocators (but will do someday)
12:58
<
whitequark >
bb-m-labs: force build --props=package=rust-core-or1k conda-lin64
12:58
<
bb-m-labs >
build forced [ETA 3m57s]
12:58
<
bb-m-labs >
I'll give a shout when the build finishes
13:02
<
whitequark >
why does it build the -14 package...
13:02
<
whitequark >
bb-m-labs: force build --props=package=rust-core-or1k conda-lin64
13:02
<
bb-m-labs >
build forced [ETA 3m57s]
13:02
<
bb-m-labs >
I'll give a shout when the build finishes
13:03
<
GitHub >
conda-recipes/master 2671bff whitequark: rust-core-or1k: bump.
13:09
<
whitequark >
bb-m-labs: force build artiq
13:09
<
bb-m-labs >
build #1473 forced
13:09
<
bb-m-labs >
I'll give a shout when the build finishes
13:17
rohitksingh has joined #m-labs
13:23
<
whitequark >
oh ffs
13:24
<
whitequark >
here's the fucking dependency
13:24
<
whitequark >
Dependency missing in current linux-64 channels: - artiq-dev 3.0.dev py_1003+gitaf20e7ed -> rust-core-or1k 1.18.0 15
13:24
<
whitequark >
how is it missing exactly?
13:36
<
whitequark >
bb-m-labs: force build --props=package=artiq-kc705-nist_clock
13:36
<
bb-m-labs >
you must provide a Builder, try 'force build [--branch=BRANCH] [--revision=REVISION] [--props=PROP1=VAL1,PROP2=VAL2...] <WHICH> <REASON>'
13:36
<
whitequark >
bb-m-labs: force build --props=package=artiq-kc705-nist_clock conda-lin64
13:36
<
bb-m-labs >
build forced [ETA 3m46s]
13:36
<
bb-m-labs >
I'll give a shout when the build finishes
13:38
<
whitequark >
bb-m-labs: force build --props=package=artiq-kc705-nist_clock artiq-board
13:38
<
bb-m-labs >
build #526 forced
13:38
<
bb-m-labs >
I'll give a shout when the build finishes
13:41
<
whitequark >
sb0: so I can quite easily do `conda install rust-core-or1k=1.18.0=15` but it fails on the buildbot.
13:42
<
whitequark >
bb-m-labs: force build --props=package=artiq-kc705-nist_clock artiq-board
13:42
<
bb-m-labs >
build #527 forced
13:42
<
bb-m-labs >
I'll give a shout when the build finishes
13:42
<
whitequark >
bb-m-labs: force build --props=package=artiq-kc705-nist_clock artiq-board
13:43
<
bb-m-labs >
build #528 forced
13:43
<
bb-m-labs >
I'll give a shout when the build finishes
13:43
<
whitequark >
bb-m-labs: force build --props=package=artiq-kc705-nist_clock artiq-board
13:43
<
bb-m-labs >
build #529 forced
13:43
<
bb-m-labs >
I'll give a shout when the build finishes
13:46
<
whitequark >
bb-m-labs: force build --props=package=artiq-kc705-nist_clock artiq-board
13:46
<
bb-m-labs >
build #530 forced
13:46
<
bb-m-labs >
I'll give a shout when the build finishes
13:49
<
whitequark >
bb-m-labs: force build --props=package=artiq-kc705-nist_clock artiq-board
13:49
<
bb-m-labs >
build #531 forced
13:49
<
bb-m-labs >
I'll give a shout when the build finishes
13:52
<
whitequark >
bb-m-labs: force build artiq
13:52
<
bb-m-labs >
build #1474 forced
13:52
<
bb-m-labs >
I'll give a shout when the build finishes
13:53
<
whitequark >
sb0: answer: sometimes conda cache can become fucked, and then it silently refuses to install packages pretending they don't exist
13:53
<
whitequark >
conda clena -i seems to fix it
14:19
ohsix has quit [Ping timeout: 260 seconds]
14:48
<
whitequark >
oh
*wonderful*
14:48
<
whitequark >
another conda issue
14:51
bb-m-labs has quit [Quit: buildmaster reconfigured: bot disconnecting]
14:51
<
GitHub171 >
buildbot-config/master 40bf335 whitequark: Bring back our homegrown cache cleanup too, for good measure.
14:51
bb-m-labs has joined #m-labs
14:55
<
whitequark >
bb-m-labs: force build artiq
14:55
<
bb-m-labs >
build #1475 forced
14:55
<
bb-m-labs >
I'll give a shout when the build finishes
15:06
<
whitequark >
sb0: Error while finding module specification for 'artiq.gateware.targets.kc705_dds' (ImportError: No module named 'artiq.gateware')
15:06
<
whitequark >
what's the fix for that?
15:07
<
whitequark >
bb-m-labs: force build artiq
15:07
<
bb-m-labs >
build #1476 forced
15:07
<
bb-m-labs >
I'll give a shout when the build finishes
15:25
<
whitequark >
sb0: hang on
15:25
<
whitequark >
I think I know why that happens
15:25
<
whitequark >
Missing dependency artiq-dev, but found recipe directory, so building artiq-dev first
15:26
<
whitequark >
why does it think the dependency is missing...
15:26
bb-m-labs has quit [Quit: buildmaster reconfigured: bot disconnecting]
15:26
bb-m-labs has joined #m-labs
15:26
<
GitHub131 >
buildbot-config/master 2925a35 whitequark: Try cleaning conda indexes too, for good measure.
15:28
<
whitequark >
bb-m-labs: force build artiq
15:28
<
bb-m-labs >
build #1477 forced
15:28
<
bb-m-labs >
I'll give a shout when the build finishes
16:17
klickverbot has joined #m-labs
16:23
<
whitequark >
sb0: I dunno.
16:23
<
whitequark >
let's just rollback conda-build.
16:23
<
whitequark >
I think the older version had less bugs and they were less bad. I have no idea what to do about these.
16:38
<
sb0 >
whitequark, but then conda will install artiq-dev=3.0 for the release-2 build.
16:40
<
sb0 >
they have a 2.1.10, could also try upgrading it. though they rarely fix bugs.
17:02
mumptai has joined #m-labs
17:08
klickverbot has quit [Ping timeout: 258 seconds]
17:16
klickverbot has joined #m-labs
17:29
klickverbot has quit [Read error: Connection reset by peer]
17:29
klickverbot has joined #m-labs
18:15
rohitksingh has quit [Quit: Leaving.]
19:20
ohsix has joined #m-labs
19:24
siruf_ has joined #m-labs
19:52
klickverbot has quit [Ping timeout: 240 seconds]
19:58
klickverbot has joined #m-labs
22:27
mumptai has quit [Quit: Verlassend]