Time |
Nick |
Message |
21:06 |
meeting |
Meeting started Mon Nov 14 21:06:56 2011 UTC. The chair is Cerlyn. Information about MeetBot at http://wiki.debian.org/MeetBot. |
21:06 |
|
Useful Commands: #action #agreed #help #info #idea #link #topic #endmeeting |
21:07 |
Cerlyn |
#meetingtopic OLPC 1.75 triage meeting 2011-11-14 |
21:07 |
|
#chair martin_xsa |
21:07 |
meeting |
Current chairs: Cerlyn martin_xsa |
21:07 |
|
cjb <cjb!~cjb 2001:4830:1171:0:224:d7ff:fe18:2a20> has joined #olpc-meeting |
21:08 |
martin_xsa |
don't sit on me |
21:08 |
|
dilinger <dilinger!~dilinger lunge.queued.net> has joined #olpc-meeting |
21:08 |
martin_xsa |
ok - agenda is to go quickly over dev.laptop.org/1.75 |
21:08 |
|
yama <yama!~yama 124-171-33-11.dyn.iinet.net.au> has joined #olpc-meeting |
21:08 |
|
yama has quit IRC |
21:08 |
|
yama <yama!~yama ubuntu/member/yama> has joined #olpc-meeting |
21:09 |
martin_xsa |
blockers and high pri items, and figure out if we need to file additional tasks |
21:09 |
|
it's our chance to make sure we don't forget about stuff |
21:10 |
|
so #10835 Implement power management code for MMP2 |
21:10 |
Quozl |
there's 115 tickets there, it will be hard not to forget about stuff. |
21:10 |
martin_xsa |
from smithbone's email, it sounds like I could stop using wmb's branch |
21:11 |
Cerlyn |
#topic blockers and high pri items |
21:11 |
martin_xsa |
dilinger: does arm-3.0-wip s/r reasonably well? |
21:11 |
smithbone |
martin_xsa: I hope so. I defer to dilinger's decision. |
21:11 |
dilinger |
martin_xsa: i hope so. :) |
21:11 |
smithbone |
The kernel he gave me ran on our bench on a 201 for 3+ days. |
21:12 |
dilinger |
i tested a few variations on that kernel, but not that kernel exactly |
21:12 |
martin_xsa |
ok, then I'll build from there. |
21:12 |
dilinger |
regardless, i'm missing the point of testing the wmb branch |
21:12 |
martin_xsa |
any further questions for marvell? or code we want released? |
21:13 |
dilinger |
there are pending questions for marvell that they haven't answered yet (about the tlb flushes) |
21:13 |
martin_xsa |
agreed - any questinos not asked yet? :-) |
21:13 |
|
question_queue_flush() |
21:14 |
dilinger |
not at the moment |
21:14 |
|
i tested their exact sram code over the weekend |
21:14 |
martin_xsa |
#10893 XO-1.75 - runin tests |
21:14 |
dilinger |
on 2 machines, i reached 10,000 cycles |
21:14 |
|
the other 2 machines crashed early on |
21:14 |
martin_xsa |
dilinger: hmm, interesting, and that's what's in -wip, so we'll see more reports then. |
21:15 |
|
right - moving on to #10893 XO-1.75 - runin tests |
21:15 |
dilinger |
martin_xsa: nah, what's in -wip is different |
21:15 |
|
-wip adds their TLB stuff, ignores some of their other changes, and keeps mitch's delays |
21:16 |
martin_xsa |
ok - I'll trust you to pick the right mix |
21:16 |
|
:-) |
21:16 |
dilinger |
it's a.. how do you say.. work in progress? ;) |
21:16 |
martin_xsa |
Quozl: for the runin tests, you have 0.16, does it include thermal? |
21:17 |
Quozl |
martin_xsa: cforth sets up thermal watchdog. openfirmware and linux do not dislodge the configuration. olpc-runin-tests 0.16.0 provides support for CPU temperature logging and (thermal) excursion halt. |
21:19 |
martin_xsa |
Quozl: right, so does 0.16 include the needed bits to talk to OFW (sdkit) and invoke the thermal tests? |
21:19 |
|
Blaketh <Blaketh!nullidentd c-69-138-76-250.hsd1.md.comcast.net> has joined #olpc-meeting |
21:19 |
Quozl |
martin_xsa: runin has no part in thermal watchdog behaviour. it is monitoring, and forcing a fail if temperature is above a characterised limit, but we have not characterised that limit yet, so it may be insensitive. |
21:20 |
|
martin_xsa: olpc-runin-tests 0.16.0 provides support for CPU temperature logging and (thermal) excursion halt. you asked how it is implemented. it uses sdkit, which is a userland build of ofw, to read the temperature sensor, because the kernel hasn't got a driver for it yet. |
21:20 |
|
martin_xsa: there is no invokation of thermal tests in ofw. |
21:20 |
|
(there are no thermal tests in ofw). |
21:21 |
martin_xsa |
Quozl: it's ok, I'm asking whether the task is done. Of course it is better to get the description right :-)) |
21:21 |
|
so it's reasonable to set this to "add to build" ? |
21:21 |
Quozl |
martin_xsa: i lack a specification to check against, sorry. all i'm going on is task list from quanta. |
21:22 |
|
martin_xsa: no, it is not reasonable to set #10893 to add-to-build, because #10893 is a metaticket, but #10954 is add-to-build, and that's where temperature monitoring for kernel is tracked. |
21:22 |
martin_xsa |
great, I hadn't seen 10954 |
21:24 |
|
right, and it is add to build. great. |
21:25 |
|
thanks. I'll add it to the build ;-) |
21:25 |
|
#11369 XO-1.75 USB stops working after suspend/resume -- |
21:25 |
|
I guess nobody is working on this one yet -- probably a good candidate after the core S/R is considered stable |
21:25 |
dilinger |
i haven't looked at that in the past week |
21:26 |
|
right |
21:26 |
martin_xsa |
#11379 PS/2 mouse, keyboard wedged after suspend/resume cycle |
21:26 |
|
pgf: I've downgraded to high this one since we have your workaround |
21:27 |
|
are there good reasons to block on having the root-cause fix? |
21:27 |
|
or a better fix? |
21:27 |
pgf |
not currently. |
21:27 |
Quozl |
does this prevent a user from typing ahead while the system is suspended? |
21:28 |
pgf |
? you can't do that now. |
21:28 |
|
in any case, there's a separate bug for first-keystroke-lost |
21:29 |
martin_xsa |
#11401 ^ |
21:29 |
|
#11395 - Implement Linux Suspend/Resume on MMP2 |
21:30 |
|
already discussed |
21:30 |
|
#11397 DCON freeze not working under X |
21:30 |
Quozl |
really? but this ticket covers X, VT, libertas as well. |
21:31 |
martin_xsa |
jnettlet has pushed a patch to -wip, and a new dovefb rpm, they'll get into the next build |
21:31 |
|
sets it to add-to-build |
21:32 |
|
#11400 Date resets to 1970 each boot |
21:32 |
|
that looks like a regression over earlier kernels |
21:32 |
pgf |
it broke when we enabled the 2nd rtc, i think |
21:33 |
|
there's a config option to tell the kernel which rtc to prefer. we should try using that to select the external rtc. |
21:33 |
martin_xsa |
cjb: do you think silbe's fix is the right one? we can add it to the defconfig |
21:33 |
|
pgf: is rtc1 the external one? |
21:33 |
pgf |
the internal rtc doesn't actually need to have the correct time. |
21:33 |
|
it must be, since rtc0 is internal. :-) |
21:33 |
Quozl |
is the discovery order predictable? |
21:33 |
dilinger |
pgf: hm. is there a runtime option, or is that compile-time only? |
21:33 |
pgf |
all i know about is the config option. |
21:34 |
cjb |
martin_xsa: if it works, it's a good idea; it involves having one of the two RTCs not knowing what the time is |
21:34 |
Quozl |
#10899, is the only design we have at the moment. |
21:35 |
martin_xsa |
cjb: looks like the ext one remembers, so if rtc1 is reliably the external one, we're go. |
21:35 |
cjb |
yes |
21:36 |
Quozl |
test that it doesn't break rtcwake. |
21:36 |
pgf |
it will, indeed, break rtcwake. |
21:36 |
Quozl |
we need rtcwake, don't we? |
21:36 |
martin_xsa |
pgf: what? |
21:36 |
pgf |
we'll have select the correct device on the rtcwake commandline. |
21:36 |
Quozl |
then we'll break runin. |
21:36 |
pgf |
oh -- not if the alarming one is rtc0 |
21:37 |
Quozl |
is the discovery order predictable? |
21:37 |
martin_xsa |
that's very interesting, the kernel cmdline says CONFIG_RTC_HCTOSYS_DEVICE |
21:37 |
pgf |
we may break hwclock. |
21:37 |
dilinger |
can we have an init script determine which rtc has the correct time (assuming it'll never be 1970 again) and set the remaining RTC devices accordingly? |
21:37 |
martin_xsa |
pgf: I think we _fix_ hwclock with this :-) |
21:37 |
pgf |
okay, maybe. |
21:38 |
martin_xsa |
dilinger: from everything I hear here, rtc0 has zero chance of being right |
21:38 |
|
it gets powered off with the SoC |
21:38 |
Quozl |
dilinger: thus trusting openfirmware never to init rtc0. |
21:39 |
|
openfirmware does have code that uses rtc0 for alarms, but i don't know when it is called. |
21:39 |
martin_xsa |
ok - sounds like the right fix, I'll apply it, and we'll see if anything breaks. if it does, we'll handle the breakage. |
21:39 |
dilinger |
martin_xsa: that assumes rtc0 is always rtc0 |
21:40 |
martin_xsa |
but the CONFIG_ variable seems to imply that it'll DTRT and not mess with the alarm clock |
21:40 |
Quozl |
it merely needs to be tested before being committed. ;-} |
21:40 |
martin_xsa |
dilinger: that's true -- Quozl, yes, I'll get that done. I'll test it over a few reboots, removing power along the way. |
21:41 |
|
and testing rtcwake, etc |
21:41 |
|
any interesting testing idea, add to ticket :-) |
21:41 |
|
#11401 keystroke which causes resume is lost |
21:42 |
dilinger |
i don't think the device ordering is in any ways guaranteed. as it happens now, the platform device for rtc-mmp gets created early, but that won't necessarily happen when that code is upstream |
21:42 |
pgf |
rtcwake has precious few clients, so if it needs a "-d rtc1" to work, that'd be okay. |
21:42 |
martin_xsa |
dilinger: is that viable to name via udev? or am I on drugs? |
21:43 |
|
CanoeBerry_ has quit IRC |
21:43 |
Quozl |
i propose "rtc_volatile" for the internal one. ;-) |
21:43 |
martin_xsa |
can we fit spawn_from_hell in there too? |
21:43 |
dilinger |
martin_xsa: yeah, i think udev should be able to control the name |
21:44 |
pgf |
they have different names in the sysfs tree already |
21:44 |
Quozl |
pgf: you said in #10899 that the one outside doesn't have alarms ... that's changed? |
21:44 |
pgf |
no |
21:44 |
|
not that i know of |
21:44 |
Quozl |
pgf: does rtcwake use alarms? |
21:44 |
pgf |
certainly |
21:45 |
martin_xsa |
dilinger: tjem we have to sync the defconfig values with the udev values... |
21:45 |
|
then |
21:45 |
|
this will be a lot of fun |
21:45 |
pgf |
this doesn't all have to be solved now. if the order is consistent today, run with it. |
21:46 |
|
if it changes when we upstream, we can deal then. |
21:46 |
martin_xsa |
good point |
21:46 |
|
erikos has left #olpc-meeting |
21:46 |
martin_xsa |
in any case, I'll explore it a bit |
21:46 |
|
Next! #11401 keystroke which causes resume is lost |
21:46 |
pgf |
no progress |
21:47 |
|
we have some ideas, but don't know if any of them will fly. |
21:47 |
martin_xsa |
is this a blocker for this stage? |
21:48 |
Quozl |
i don't know the blocker criteria. |
21:48 |
pgf |
it's not a great user experience, as you'll find out when aggressive s/r comes to your laptop. |
21:48 |
|
what's "this stage"? |
21:48 |
martin_xsa |
current blocker criteria is: it blocks C2 runin |
21:48 |
Quozl |
i don't know if quanta will whinge about it. |
21:48 |
|
it does not block C2 runin. |
21:48 |
pgf |
right |
21:48 |
martin_xsa |
exactly -- Quanta, runin, tests on mfg line |
21:49 |
Quozl |
it may cause Quanta's QA team to raise it because it is different, but not if we don't tell them. ;-) |
21:49 |
martin_xsa |
I'll demote it to high -- naturally it's a blocker to a shipping OS |
21:49 |
|
pgf: we don't think it's got a hw component, right? |
21:50 |
Quozl |
#11369 is not a blocker for C2 runin, since USB is not tested by runin. |
21:50 |
pgf |
not specifically, no. not any more than every other s/r related bug. |
21:51 |
Quozl |
#11397 is not a blocker for C2 runin, since nobody is meant to watch the screen. |
21:51 |
|
#11400 would impact C2 runin file datestamps, needs fixing, but the C2 runin tests would pass. |
21:52 |
|
#11409 is not a blocker for C2 runin, since runin does not test headphone jack. |
21:53 |
martin_xsa |
Quozl: quanta does look at the units during runin, and will report oddities #11397 |
21:55 |
|
cjb has quit IRC |
21:55 |
martin_xsa |
and Quanta does test things and performs basic tests on machines after runin, so #11369 is certain to hold things up if we enable s/r in powerd |
21:55 |
Quozl |
Of the "high" tickets, the following may impact C2 runin ... #11237, #11357, #11408, #11413, #11416, #11430 (if Quanta don't like it), #11438). |
21:55 |
martin_xsa |
ok - we are going over them :-) |
21:56 |
|
it's not the onlything we want to know |
21:56 |
|
cjb <cjb!~cjb 2001:4830:1171:0:224:d7ff:fe18:2a20> has joined #olpc-meeting |
21:57 |
martin_xsa |
pgf: thanks |
21:57 |
|
I'll skip audio for now, saadia couldn't join the meeting |
21:58 |
|
Cerlyn: you reported that #11196 some 1.75 touchpads "wander" in ebook mode |
21:58 |
|
still happens a bit? |
21:58 |
Cerlyn |
a bit but not much with local testing |
21:58 |
|
and that's if you are moving something behind the touchpad in ebook mode |
21:59 |
|
or moving the XO to effectively do the same |
21:59 |
pgf |
that's significantly different than originally, right? it used to wander as soon as you closed the screen against the keyboard. |
21:59 |
martin_xsa |
ok - anyway, that one's on Quanta's side, and we have olpc-kbdshim masking it |
22:00 |
pgf |
it's really a vendor acceptance issue. |
22:00 |
Cerlyn |
pgf: I think that's environmental conditions or potentially sample dependent |
22:00 |
martin_xsa |
I am not too worried about it, -- agreed with pgf |
22:00 |
|
Next: #11237 Video driver hang on XO-1.75 os4 |
22:00 |
|
I agree with Quozl that's potentially a blocker |
22:00 |
pgf |
martin_xsa: i just don't want quanta to say "not a problem" because we've worked around the problem in s/w. |
22:01 |
martin_xsa |
pgf: absolutely -- we want the tp to be quiet so we don't wake up |
22:01 |
|
pgf: if the vendor gives us a userland command to shut it up... |
22:02 |
pgf |
martin_xsa: the already suggested flow-controlling the touchpad in h/w. |
22:02 |
|
s/the/they |
22:03 |
|
it's not out of the question, i guess. |
22:03 |
martin_xsa |
pgf: but the window on hw changes is closed, unless we have things wired up so that we can do it? |
22:03 |
pgf |
ps2 flow control is part of normal operation. no physical h/w changes needed. |
22:03 |
|
(i shouldn't have said "in h/w" above.) |
22:04 |
|
(i meant "at the h/w level") |
22:04 |
martin_xsa |
pgf: ah, so that'd be on the SP |
22:04 |
|
Quozl: #11237 not a blocker, it triggers with an accelerated video option we aren't using now |
22:04 |
pgf |
martin_xsa: or in the EC on the other laptops. remember that they'll use that pad everywhere. |
22:05 |
martin_xsa |
pgf: yep |
22:05 |
cjb |
martin_xsa: are we shipping without any gpu acceleration, then? |
22:06 |
martin_xsa |
cjb: you're parachuting into it :-) |
22:06 |
cjb |
hm? |
22:06 |
pgf |
the flow-control sol'n isn't as easy on the 1 and 1.5, because the EC doesn't have the "i'm an ebook" signal. |
22:07 |
martin_xsa |
cjb: current criteria is for C2 build & runin, and #11237 just doesn't meed the blocker crit |
22:07 |
|
jon promises a fix soon, though |
22:07 |
|
cjb: any progress on #11353 SD Card cannot detected in OS881? I did some basic tests... |
22:08 |
cjb |
martin_xsa: does suspend/resume crash if UNSAFE_RESUME's off? |
22:08 |
martin_xsa |
cjb: I didn't test /that/ :-) |
22:09 |
Quozl |
pgf: i've just checked the current hardware call issues list, and there is no current entry for CL2 that covers #11196 wander, so please raise it with wad? |
22:09 |
cjb |
martin_xsa: that would decide whether it's blocker. |
22:10 |
martin_xsa |
cjb: it crashes |
22:10 |
cjb |
ok |
22:10 |
pgf |
Quozl: martin_xsa was driving that process with quanta and the vendor. they were going to provide new firmware or somesuch. i assume martin's still on top of it, and won't accept the new h/w until it passes our criteria. |
22:10 |
martin_xsa |
cjb: last week you said you'd investigate -- any progress on that? |
22:10 |
cjb |
martin_xsa: no, been too busy with XO-3 |
22:10 |
Quozl |
pgf: 'k. |
22:11 |
martin_xsa |
pgf, Quozl - yep, it's in the quanta confcall topic |
22:11 |
|
topic/task list |
22:11 |
Quozl |
martin_xsa: i must have missed it, i just went through it. |
22:12 |
cjb |
martin_xsa: what's your deadline for getting the blockers fixed? |
22:12 |
Quozl |
(several closed items). |
22:12 |
|
i used the tracking list dated 2011-11-10, you might have another one. |
22:12 |
martin_xsa |
cjb: right now we're working towards C2 build, so blockers are defined as to whether they block C2 |
22:12 |
cjb |
right, I know |
22:13 |
martin_xsa |
somethings are high pri just because they just are, and affect us beyond C2 |
22:13 |
Quozl |
now that i know the blocking criteria, things are clearer to me. it's good. |
22:13 |
martin_xsa |
I want to have the OS img for C2 1st of Dec, so closing date is Nov 30th |
22:14 |
|
that's laid out in an email to techteam and devel, with titled 11.3.1 timeline or similar |
22:15 |
|
cjb: I assume xo-3 is gonna keep you busy++ |
22:15 |
cjb |
Okay, thanks. I'm ambivalent on whether to commit to this because Scott's currently blocked on me figuring webgl out. At least there are two weeks rather than a couple of days for it, though. |
22:15 |
|
(to fix the SD bug, whatever it is) |
22:15 |
martin_xsa |
cjb: ok |
22:16 |
|
understood |
22:16 |
cjb |
thanks. let's revisit next week, things should be less crazy then. |
22:16 |
martin_xsa |
ok |
22:16 |
|
next #11357 Boot freezing on third clock dot |
22:17 |
|
as Quozl, this is a candidate blocker |
22:17 |
Quozl |
i regularly see boot freezes on b1 and c1 units here, when i get to try linux. |
22:17 |
martin_xsa |
yep- I see it lots too |
22:18 |
Quozl |
they are either third-clock-dot or scsi_wait_scan. |
22:18 |
pgf |
are you distinguishing between the two, or saying they're the same thing? |
22:18 |
|
(i think the latter) |
22:18 |
dilinger |
huh. i don't think i've ever seen that |
22:19 |
Cerlyn |
pgf: I believe they are the same thing; and technically the modprobe is blocked on something else and it may not be the module's fault |
22:19 |
Quozl |
i don't know which is which, sometimes i'm looking at the screen, sometimes i'm not, sometimes i've booted from ok, sometimes i've got a USB to ethernet dongle involved, i just haven't fully diagnosed and separated out those symptoms, i was usually busy on some ofw errand. ;-} |
22:19 |
martin_xsa |
dilinger: do you build your psmouse into the kernel? |
22:19 |
cjb |
I see it a lot on my C1, feel like I hadn't ever seen it pre-C1 |
22:19 |
dilinger |
martin_xsa: nope, i'm using the defconfig |
22:20 |
martin_xsa |
dilinger: that must be a pain :-/ |
22:20 |
pgf |
dilinger: do you have modules present? |
22:20 |
dilinger |
though i was using a non-modular kernel for the majority of my s/r testing, up until a few weeks ago. so maybe that's why |
22:21 |
|
martin_xsa: it's not too bad; since i've been working on the sram code non-stop, every rebuild only affects vmlinuz |
22:21 |
martin_xsa |
maybe built-in psmouse changes the order of things just enough that we sidestep this one :-P |
22:21 |
|
pbrobinson <pbrobinson!~pbrobinso 2001:41d0:2:923e::> has joined #olpc-meeting |
22:23 |
martin_xsa |
I'll try to see if dsd has any more debugging ideas |
22:23 |
|
Next #11359 Psmouse kernel panic on shutdown |
22:25 |
|
This doesn't break runin, but is damn weird - maybe we're trying to talk to AVC touchpads too fast. |
22:25 |
Quozl |
#11359, no impact to C2 runin phase, since the power button is not pressed twice. |
22:25 |
Cerlyn |
pgf and further testing suggests it may also happen with S/R cycles |
22:26 |
Quozl |
#11359, if it happens with suspend only with keyboard or mouse input, then it shouldn't be a concern for runin. |
22:26 |
Cerlyn |
does not seem to be the most common S/R failure here though |
22:27 |
martin_xsa |
11359 is a symptom -- at shutdown time -- of failure to init AVC touchpads |
22:28 |
pgf |
i wouldn't be surprised if it's gone from s/r, since i disabled mouse resets across s/r. |
22:28 |
|
(pushed today) |
22:28 |
martin_xsa |
there's ~4 bugs reporting the symptoms, root cause seems to be squarely in psmouse init process |
22:28 |
|
ah, good point, maybe now we wiggle it less |
22:29 |
Cerlyn |
2 test kernels used for S/R testing have it; once with os10 on the first S/R cycle |
22:29 |
|
at least in my S/R log library |
22:29 |
martin_xsa |
let's run quickly over the remaining high bugs -- to see if we have candidates for blockers. Quozl identified a few |
22:30 |
|
#11396 Implement Linux support for wake-on-lan on MMP2 -- this is waiting for S/R stability |
22:31 |
|
#11399 [CL2]Press enter many times in write activity and system will hang up -- not a blocker but important -- I hope jnettlet tackles it soon |
22:33 |
|
#11406 powerd needs to be ported to new "wakeup events queue" kernel backend -- not a blocker -- but important, can we separate out the kernel tasks? |
22:33 |
pgf |
it's done. modulo bug fixes. |
22:34 |
martin_xsa |
pgf: oh! great |
22:34 |
pgf |
and modulo missing parts, like libertas. |
22:34 |
martin_xsa |
that's the recent string of commits on -wip? |
22:34 |
pgf |
parly. half were that, half were the OLS support. |
22:34 |
martin_xsa |
yeah, that's what I meant |
22:35 |
pgf |
also modulo getting it into a build, which i've failed miserably at today. see mail. |
22:35 |
martin_xsa |
for powerd? |
22:36 |
|
ok - I'll look into it, also tmw morning, if you're not around on irc, I'll give you a step-by-step of my steps and rationale |
22:36 |
|
and the img build will happen tmw anyway :-) |
22:37 |
pgf |
i'm sure it's easy if one does it frequently. i'd appreciate a step-by-step. |
22:37 |
|
my officemates would appreciate me getting that help as well, i'm sure. |
22:37 |
martin_xsa |
I don't do it frequently, but I've done it a few times... so it'll be very step by step. |
22:37 |
|
:-) |
22:40 |
|
updated 11408 to block S/R, and be a blocker too |
22:40 |
|
same with #11413 XO-1.75 libertas crash on resume |
22:42 |
|
actually, #11413 doesn't quite make it to a blocker |
22:43 |
dilinger |
the libertas-crash-on-resume is high on my todo list, post-ram stabilization |
22:43 |
martin_xsa |
excellent! |
22:43 |
|
it is odd -- it only crashes if associated |
22:44 |
dilinger |
if you leave NM running, even if you're not telling it to associate, it'll trigger the bug quickly |
22:45 |
|
jnettlet <jnettlet!~jnettlet 96.49.157.58> has joined #olpc-meeting |
22:45 |
martin_xsa |
#11416 C1 XO 1.75s may not wake up due to RTC from suspend -- marked as blocker |
22:46 |
|
#11430 mmp-camera output corrupt after suspend/resume -- hesitant about his one, it's important but won't delay C2 on it |
22:46 |
|
jon corbet is working on it |
22:47 |
dilinger |
has the not-wake-up-from-suspend issue happened since pgf's keyboard workaround went in? |
22:49 |
Cerlyn |
when was that committed? |
22:49 |
martin_xsa |
it's in os10 |
22:49 |
|
and any kernels we tested on top of os10 |
22:50 |
Cerlyn |
I don't think it's happened since |
22:50 |
|
but I may have woken some up this morning |
22:51 |
dilinger |
it may have been a timing problem that was fixed by pgf's commits, then. the rtc alarm was firing, but due to keyboard/touchpad bugs the system was delayed in getting to sleep.. by the time the system had gone to sleep, the alarm had already fired |
22:52 |
|
if anyone sees it with os10+, please update the bug w/ dmesg output or something |
22:52 |
martin_xsa |
dilinger: yeah -- though that can still happen with very short rtcwake calls |
22:53 |
pgf |
dilinger: good thought. i can't say for sure that that's not what i've seen. |
22:53 |
martin_xsa |
methink we should have a guard around rtcwake invokations, that ignores irequests to sleep shorter than a minimum time (that being a safe amount of time to down cleanly) |
22:54 |
pgf |
the new kernel PM infra supposedly makes it possible to prevent the wakeup-happens-while-suspending race. |
22:54 |
|
i haven't tried implementing it yet, but i'm working toward it. |
22:54 |
|
martin_xsa: that won't help, if the system takes longer to get to sleep than just >< this much. |
22:55 |
martin_xsa |
pgf: you mean new kernel pm infra we are not currently using? |
22:56 |
dilinger |
martin_xsa: i know mitch was a fan of 2s alarms; i tried to get people to use 5s wakeup alarms. i think 5s is a good amount of time; if we delay that long on the way down, something is definitely wrong |
22:56 |
martin_xsa |
linux 3.1 or 3.2? |
22:56 |
|
dilinger: sure, and as we debug the process, we'll get it shorter |
22:56 |
pgf |
martin_xsa: i think our current kernel has the necessary support. |
22:57 |
martin_xsa |
pgf: but we hit this issue quite a bit. if the kernel infra "prevents" this from happening... |
22:57 |
|
maybe that word doesn't mean what the implementor of this infra thinks it means... |
22:58 |
|
anyway, later on we'll know now long it takes us to quiesce our system |
22:58 |
pgf |
martin_xsa: i just said "i haven't tried implementing it yet, but i'm working toward it." |
22:58 |
martin_xsa |
ah, sorry, there's a way to wire it up. oops, sorry. |
23:01 |
|
ok - reading down the trac list. I don't see anything that looks likely to block us towards C2 |
23:03 |
|
pgf, dilinger, smithbone, jnettlet, do you have items you'd like to make blockers or discuss? |
23:03 |
|
and Quozl :-) |
23:03 |
jnettlet |
nothing new nope. |
23:04 |
smithbone |
martin_xsa: nope |
23:04 |
pgf |
nope |
23:05 |
martin_xsa |
great -- now all we have to do is fix all the bugs discussed beforen next monday. |
23:05 |
|
meeting adjourned ;-) |
23:05 |
meeting |
martin_xsa: Error: "adjourned" is not a valid command. |
23:05 |
Cerlyn |
#endmeeting |
23:05 |
meeting |
Meeting ended Mon Nov 14 23:05:25 2011 UTC. Information about MeetBot at http://wiki.debian.org/MeetBot. (v 0.1.4) |
23:05 |
|
Minutes: http://meeting.sugarlabs.org/o[…]-14T21:06:56.html |
23:05 |
|
Log: http://meeting.sugarlabs.org/o[…]11-11-14T21:06:56 |