IRC log of #maemo-ssu for Sunday, 2011-12-04

*** mase_76 has joined #maemo-ssu02:14
*** mase76 has quit IRC02:17
*** scoobert1on has quit IRC02:23
*** mase76 has joined #maemo-ssu02:24
*** mase_76 has quit IRC02:27
*** mase_76 has joined #maemo-ssu02:39
*** mase76 has quit IRC02:43
*** mase76 has joined #maemo-ssu02:50
*** arcean has quit IRC02:50
*** jonwil has joined #maemo-ssu02:50
*** mase_76 has quit IRC02:53
jonwilSicelo: ping02:55
jonwilmohammadag: ping02:55
*** mase76 has quit IRC03:04
*** merlin_1991 has joined #maemo-ssu03:54
*** merlin1991 has quit IRC03:57
*** Jade has quit IRC04:18
*** jonwil has quit IRC04:46
*** amiconn has quit IRC05:47
*** amiconn_ has joined #maemo-ssu05:47
*** amiconn_ is now known as amiconn05:47
*** nox- has quit IRC05:48
*** DocScrutinizer has quit IRC07:29
*** DocScrutinizer has joined #maemo-ssu07:29
*** jonwil has joined #maemo-ssu09:16
freemangordonDocScrutinizer, so far my theory that thumb problem is because of buggy gcc seems true. gcc 4.6.2 recompiled Qt seems rock stable so far. I think to upload a link to .debs so brave souls could test.10:30
freemangordonsize difference is 7.7 against 10.5 MB for libQtGui and 11.9 against 17.7 MB for libQtWebkit.10:41
freemangordonand that's with -O3, which results in a bigger binaries that -O2 or -Os10:41
*** mirandir has joined #maemo-ssu12:02
*** mase76 has joined #maemo-ssu12:39
*** psycho_oreos has quit IRC12:45
*** Pali has joined #maemo-ssu13:05
*** mirandir has left #maemo-ssu13:12
*** NIN101 has joined #maemo-ssu13:39
*** arcean has joined #maemo-ssu13:54
Sicelojonwil: pong14:04
jonwilnow I cant remember why I wanted to talk to you...14:05
Sicelomaybe reinstall off libconnui..14:05
jonwilyeah possible14:05
Siceloweird. widget was working fine the whole time. just started doing that after a reboot.14:06
Sicelobtw, i have gdb on device now (although, of course, i don't have a clue how to use it)14:07
Sicelocame with MADDE14:07
jonwilso you are having no issues whatsoever with the widget now?14:09
Siceloyeah, fine for now14:10
*** Jade has joined #maemo-ssu14:12
*** Jade has joined #maemo-ssu14:12
*** m0use has quit IRC15:04
*** mase_76 has joined #maemo-ssu15:16
*** m0use has joined #maemo-ssu15:27
*** Milhouse has quit IRC15:42
DocScrutinizerMEH15:42
DocScrutinizerthanks for killing mermo15:43
*** Milhouse has joined #maemo-ssu15:43
RST38h???15:43
RST38hHave I missed something, again?15:43
DocScrutinizer[2011-12-04 09:30:57] <freemangordon> DocScrutinizer, so far my theory that thumb problem is because of buggy gcc seems true. gcc 4.6.2 recompiled Qt seems rock stable so far. I think to upload a link to .debs so brave souls could test.15:45
*** cyteen has quit IRC15:46
*** X-Fade has quit IRC15:46
*** arcean has quit IRC15:46
*** jonwil has quit IRC15:46
*** amiconn has quit IRC15:46
*** freemangordon has quit IRC15:46
*** ruskie has quit IRC15:46
*** regawyte has quit IRC15:46
*** xnt14 has quit IRC15:46
*** MohammadAG has quit IRC15:46
*** DocScrutinizer51 has quit IRC15:47
*** merlin_1991 has quit IRC15:47
*** akikhaw has quit IRC15:47
*** wmarone has quit IRC15:47
*** SpacedOut has quit IRC15:47
*** vStone has quit IRC15:47
*** NIN101 has quit IRC15:47
*** xmlich02 has quit IRC15:47
*** IronLegend has quit IRC15:47
*** luke-jr has quit IRC15:47
*** chainsawbike has quit IRC15:47
*** Milhouse has quit IRC15:47
*** Pali has quit IRC15:47
*** DocScrutinizer has quit IRC15:47
*** M4rtinK has quit IRC15:47
*** mase_76 has quit IRC15:47
*** andre__ has quit IRC15:47
*** ZogG has quit IRC15:47
*** Raimu has quit IRC15:47
*** Sicelo has quit IRC15:47
*** claw has quit IRC15:47
*** RST38h has quit IRC15:47
*** thingie has quit IRC15:47
*** peetah has quit IRC15:47
*** m0use has quit IRC15:47
*** mase76 has quit IRC15:47
*** Sc0rpius has quit IRC15:47
*** neal has quit IRC15:47
*** gregoa has quit IRC15:47
*** Lava_Croft has quit IRC15:47
*** macmaN has quit IRC15:47
*** ekze-nyan has quit IRC15:47
*** ThreeM has quit IRC15:47
*** ChanServ has quit IRC15:47
*** IronLegend has joined #maemo-ssu15:52
*** luke-jr has joined #maemo-ssu15:52
*** DocScrutinizer51 has joined #maemo-ssu15:52
*** ruskie has joined #maemo-ssu15:52
*** MohammadAG has joined #maemo-ssu15:52
*** xnt14 has joined #maemo-ssu15:52
*** regawyte has joined #maemo-ssu15:52
*** freemangordon has joined #maemo-ssu15:52
*** merlin_1991 has joined #maemo-ssu15:52
*** amiconn has joined #maemo-ssu15:52
*** jonwil has joined #maemo-ssu15:52
*** arcean has joined #maemo-ssu15:52
*** ThreeM_ has joined #maemo-ssu15:52
*** xmlich02 has joined #maemo-ssu15:52
*** vStone has joined #maemo-ssu15:52
*** SpacedOut has joined #maemo-ssu15:52
*** wmarone has joined #maemo-ssu15:52
*** ekze has joined #maemo-ssu15:52
*** akikhaw has joined #maemo-ssu15:52
*** chainsawbike has joined #maemo-ssu15:52
*** NIN101 has joined #maemo-ssu15:52
*** cyteen has joined #maemo-ssu15:52
*** X-Fade has joined #maemo-ssu15:52
*** Milhouse has joined #maemo-ssu15:52
*** Pali has joined #maemo-ssu15:52
*** DocScrutinizer has joined #maemo-ssu15:52
*** M4rtinK has joined #maemo-ssu15:52
DocScrutinizerten or twenty other guys looked into it and found thumb is riddled by SiERR, but nooo, freeman doesn't believe it. Prefers to provide thumb binaries "for test" rather than discussing the basic issue with those who should know more than he does.15:53
*** m0use has joined #maemo-ssu15:53
*** mase_76 has joined #maemo-ssu15:53
*** mase76 has joined #maemo-ssu15:53
*** andre__ has joined #maemo-ssu15:53
*** ZogG has joined #maemo-ssu15:53
*** Sc0rpius has joined #maemo-ssu15:53
*** neal has joined #maemo-ssu15:53
*** thingie has joined #maemo-ssu15:53
*** Raimu has joined #maemo-ssu15:53
*** gregoa has joined #maemo-ssu15:53
*** Lava_Croft has joined #maemo-ssu15:53
*** macmaN has joined #maemo-ssu15:53
*** Sicelo has joined #maemo-ssu15:53
*** claw has joined #maemo-ssu15:53
*** RST38h has joined #maemo-ssu15:53
*** peetah has joined #maemo-ssu15:53
DocScrutinizerfriggin netsplit15:53
DocScrutinizerten or twenty other guys looked into it and found thumb is riddled by SiERR, but nooo, freeman doesn't believe it. Prefers to provide thumb binaries "for test" rather than discussing the basic issue with those who should know more than he does.15:53
freemangordonDocScrutinizer, well I tried. And the result was - "I won't tell you why, but it does not work".15:55
freemangordonAnd I am not buyng that15:56
freemangordon*buying15:56
freemangordonAnd if thumb compiled Qt,gtk,h-d and hildon-home (i.e. the most frequently executed code pieces) does not crash on several (or tenths) of devices for a month or so I will assume that stable.15:59
RST38h lemme ask a better question: what do you get from using thumb as opposed to arm32?16:00
freemangordonmemory16:00
freemangordonand BTW it is thumb2 actually16:00
freemangordondoes not have performance penalty16:01
RST38hOk, so it takes less memory, indeed16:01
freemangordonQt libraries are about 12 MB smaller16:01
*** ChanServ has joined #maemo-ssu16:02
*** hubbard.freenode.net sets mode: +o ChanServ16:02
freemangordonmemory, cache, less swapping...16:02
RST38h[shudder] what is the total size?16:02
freemangordonwell, "size difference is 7.7 against 10.5 MB for libQtGui and 11.9 against 17.7 MB for libQtWebkit"16:02
freemangordonnot checked other libs, but that should be more than enough16:03
RST38hholy fuck16:03
RST38hDoc: Have you considered the idea that GCC's backend might have been patched with respect to that errata?16:04
freemangordonand IF it is stable and IF all libs are thumb re-compiled then you can imagine the memory gain we will have16:04
freemangordonI can think of about 20-40 MB16:05
freemangordonright now i have 55MB swap which is unusual16:06
freemangordonRST38h, which errata?16:07
freemangordonwill someone point me to some data at last, everyone is saying "that" errata.16:08
DocScrutinizerfreemangordon: I'm happy now you got no maintainer powers16:08
*** ThreeM_ is now known as ThreeM16:08
freemangordonDocScrutinizer, well seems everyone is happy, including me :D16:08
freemangordonBut I fail to see how is that related to OMAP erratas16:10
DocScrutinizerRST38h: I already spent THREE hours reading thru errata and mostly talking to people who investigated all that the *right* way by actually understanding what's the nature of the problem and if and how it might get fixed - they all tell me it *can not* get fixed for our silicon. Now you suggest I invest more of my precious time for freemangordon's "I don't buy it"? MEH16:12
DocScrutinizeractually I did this 3h fancy just for freemangordon to find the details he's been too lazy to look/ask for16:12
DocScrutinizeras I already *knew* thumb is broken and deprecated16:12
DocScrutinizeron a hw level16:13
DocScrutinizernow I have to learn it was all futile16:13
DocScrutinizerand GODDAMNIT I knew better ways to save 20MB of RAM than messing with a deprecated proven defect CPU opcode16:15
DocScrutinizere.g. locale16:16
RST38hDoc, not being able to fix something in silicon does not mean you cannot fix it in software16:16
RST38hIn this case, you fix it by patching GCC backend not to produce offending instruction sequences16:16
freemangordonDocScrutinizer, you were the one to find one errata fixed in gcc16:16
DocScrutinizerRST38h: thanks for informing me of this very new fact16:16
RST38hHave you considered this approach and checked the latest gcc changelist?16:16
RST38hDoc: Sorry for playing CptObvious =)16:17
freemangordonBTW gcc I am using was not used by "knowledgable guys"16:17
DocScrutinizerRST38h: why didn't freemangordon do that?16:17
DocScrutinizerI suggested to him to show where the known SiERR are attributed "FIXED by workaround" and I'd be interested again16:18
RST38hI have no idea. I also have no idea what the point in this conversation is.16:18
RST38hHe has made whatever testing he has made and decided that his stuff is safe.He is going to release it as part of CSSU as I understand.16:19
freemangordonno16:19
RST38hIf his stuff blows, we boo at him, he rolls it back, and everything is fine16:19
freemangordon:D16:19
freemangordon(nod)16:19
DocScrutinizerFUCK NO WAY!16:19
RST38hSo, Doc, why bitch?16:19
RST38hDoes his stuff cause N900 to explode in fiery death?16:20
DocScrutinizerCSSU is not experimental for weird "mad scientists" to do their mass studies16:20
freemangordonDocScrutinizer, i would't qualify myself as "mad scientist", why would you?16:21
freemangordonanyway, bb for now. bbl16:22
RST38hAnyway, gentlemen, before I leave this place for RL, why not stop this endless argument and do something pleasant for a change, like sacrificing a hamster to the Tentacled One?16:22
DocScrutinizerbecause you don't listen to sound reasons and arguments, rather do exactly what's been explained to your (and generally accepted in real science) to be a nogo: basing assumptions on random testing ignoring the need of properly understanding the known available facts16:23
DocScrutinizer[2011-12-03 16:38:46] <DocScrutinizer> even when he came up with a whole Qt thumb build that uses 40% less mem footprint and runs fine on his device and system for days, we wouldn't want to touch it with a 10 ft pole16:26
DocScrutinizer[2011-12-03 16:39:19] <DocScrutinizer> unless he came up with a *comprehensive* list of *all* silicon errata in 3430, and analyzed each single one of them and explained how his workaround is proven to deal with it so it won't cause anymore problems16:26
DocScrutinizerRST38h: which is exactly the "look into gcc changelog" you suggested16:27
DocScrutinizeror at least includes that action, not sure if it suffices16:27
DocScrutinizereven if the SiERR were listed as "FIXED with version XY of gcc" we still had to check if that patch actually got verified or if just another mr freeman pushed those and nobody actually checked if they are effective16:30
DocScrutinizerit's not exactly easy to verify a gcc workaround fix for a SiERR like "problem: when conditional shortjump after flushing cache has a destination that's exactly the start of the N+2 4k page of code, this instruction will fail if it's a conditional OPcode that has negative result on condition. Workaround: have a redirect stub in N+3 page to jump there first, and from there jump to final destination"16:35
DocScrutinizer(example made up)16:37
DocScrutinizerbut real SiERR are of same complexity16:37
DocScrutinizernow TEST this to verify your compiler patch A) actually got implemented in a correct way that always works, and B) that the fix/workaround regarding CPU operation actually is effective16:39
DocScrutinizerodds are you create MB of code and the offending situation isn't ever occurring in it16:39
DocScrutinizereven if it is, you never know how often and when the execution time conditions are met so the potentially dangerous code actually triggers the error16:41
DocScrutinizerand the icing on top: NOBODY literally NOBODY will be able to tell if it's been your defective "fix" that failed when eventually a user's dialer or H-D or whatever reboots out of a sudden16:43
DocScrutinizerresult: general image of CSSU/maemo becomes "unstable crap"16:44
DocScrutinizerand NO WAY we are going to include sth like that into CSSU, based on a WFM basis16:45
DocScrutinizerafter Nokia and meego chief developers had extensive looks into it and all considered it broken or at very least too risky16:46
DocScrutinizerpretty much reminds me at futile debates with OC-blockheads16:46
DocScrutinizerRST38h: as well we could optimize out all mutex and spinlocks, claiming "it works for me" as you maybe never run into the races that may occur - but *somebody* will, pretty soon. And nobody would complain to us about "2 of the 7 reboots of last week were due to your missing spinlocks"16:48
RaimuDoing testing runs for that sort of an issue in a community where every, every new patch of every piece of software gets "battery drain started" or "random reboots started" bug reports...16:49
RaimuThat sounds like bad form incarnate.16:49
DocScrutinizerindeed, useless16:49
RaimuEspecially when many people consider the device stable if their N900 does a random totally unexpected reboot even ONCE and WEEK.16:50
RaimuThat is NOT stable.16:50
DocScrutinizeragain, exactly16:50
DocScrutinizernow count in that same freemangordon blamed me about "wasting time on useless stuff, better do something actually useful" when I tried to sort the camera-ui issue by suggesting a picker to select default camera-ui16:52
DocScrutinizerand we goddamn have other memory hogs that are WAY easier to kill, e.g locale, tracker, general preload of dog and cat16:53
RaimuThat picker thing was more along "personal opinion" lines, I can see how it sometimes feels someone isn't focusing on what's important.16:54
DocScrutinizerwhere the risks are pretty obvious and manageable, and reports we'd see when something goes mayhem are specific and easy to attribute to a particular "fix"16:55
RaimuIt's just this... bloody-minded centering on what is the desired outcome while conveniently ignoring potential hazards that irks me.16:55
DocScrutinizeryep16:55
DocScrutinizerindeed16:55
*** Sicelo_ has joined #maemo-ssu16:59
DocScrutinizerohwell, more snake oil to the masses16:59
*** Sicelo_ has quit IRC16:59
*** Sicelo_ has joined #maemo-ssu16:59
DocScrutinizerI *bet* they'll like it17:00
DocScrutinizerepitaph of maemo will be "awesome OS, awesome community, lots of brilliant apps and community driven solutions like overclocking, thumb bin size reduction, speed optimizer, zillion of applets. Could have been a success if only the hardware wasn't that crappy: slow, segfaults all the time, and CPU burns out after 4 weeks of intensive use"17:04
DocScrutinizeroooh, forgot "unbearable standby time, way too small battery"17:04
RaimuHaha. That's mean.17:31
*** andre__ has quit IRC17:32
*** andre__ has joined #maemo-ssu17:34
*** andre__ has joined #maemo-ssu17:34
*** jonwil has quit IRC17:38
*** andre__ has quit IRC17:39
*** andre__ has joined #maemo-ssu17:49
*** andre__ has joined #maemo-ssu17:49
freemangordonDocScrutinizer, Raimu, let me tell you something. First of all I would *NOT* put anything into users devices marked as STABLE on WFM basis. Second (and that is what irritates me most) is that you are trying to tell me to not try it on my device because someone else has failed. Don't give a shit how famous those guys are. I need to check it. Not that I am going to push thumb compiled Qt in17:49
freemangordonCSSU tomorrow. And *IF* it is stable *THEN* I will check in gcc changelogs or wherever what and how has been fixed. It just does not make sense to spend a week reading documentation if the binary explodes in a minute. And i won't continue that discussion furter, as it is absolutely pointless.17:49
DocScrutinizerfair enough17:51
Raimufreemangordon: I'm not stopping you. I was actually rather optimistic about it until Doc here pasted a bunch of documents and conversations that pretty much made the case for me.17:54
Raimufreemangordon: It'd be wonderful if thumb actually worked instead of borked on the n900 hardware. I merely lost my "likelihood of this happening" spirit.17:55
Raimufreemangordon: And you seem very determined - still, from what I understand this type of glitch isn't the "binary explodes in a minute" variety but rather a bunch of sneaky, hard to trace error events in the long run.17:57
freemangordonRaimu, will you please point me to those documents , seems I've missed something (sorry DocScrutinizer, shit happens). I really apologize if that is the case.17:58
DocScrutinizerjust I never was " trying to tell [you] to not try it on my device because someone else has failed."  I went OMGWTF when I read things like >> I think to upload a link to .debs so brave souls could test<< and >>He has made whatever testing he has made and decided that his stuff is safe.He is going to release it as part of CSSU as I understand. If his stuff blows, we boo at him, he rolls it back, and everything is fine <freemangordon> :D17:58
DocScrutinizer (nod)<<17:59
Raimufreemangordon, sorry. I didn't keep a backlog.17:59
RaimuMaybe DocScrutinizer has them (and some patience still) to repaste them.18:01
freemangordonDocScrutinizer, what is the problem if someone wants to try it?18:02
DocScrutinizerRST38h missed that when you're messing with Qt it will never be *your* stuff that blows, and nobody will open a ticket against this Qt thumbification18:02
DocScrutinizerI wouldn't give a f*ck about any app built with thumb (unless it's an app like CSSU-modest supposed to be mandatory on a huge number of devices)18:03
DocScrutinizermessing with Qt, HD, core components in general: whole different class of stuff18:04
DocScrutinizerthe tickets will never reach you, as nobody can tell acme-app actually broke because of a thumbified lib or whatever shared resource18:05
DocScrutinizerand I don't even see any willingness to think about concepts to actually tackle *that* issue (validation in field, proper backtrace for occasional errors happening, testing strategy at large. WFM is NO testing strategy)18:08
DocScrutinizer"let's assume there is no problem (ignoring the fact there are lots of indications there actually is one), and see what we get when we simply use it as if it always worked" is no feasible strategy either18:11
DocScrutinizeryour above sketched approach of first testing locally, then go to the *changelogs* - way better a strategy18:11
freemangordonDocScrutinizer, it's been less than 24 hours I've been able to set-up the new toolchain and you are asking for strategy.18:12
freemangordonIf i have a single crash in a week then I will say here " I was proven wrong, I am stupid". But if there is no evidence of that I will google for as much info as possible.18:13
DocScrutinizersounds sane18:14
DocScrutinizerno objections18:14
freemangordonThat was my point all the time18:14
DocScrutinizersorry for getting you wrong then18:14
freemangordonSee, I don't have any plans for pushing that into CSSU at least for the next 3 months. And I will try to push it in CSSU  *ONLY* and *ONLY IF* there is no evidence that thumb causes crashes, i.e. devices running for 2-3 week without reboot and without a single application crash, erratas fixed, etc. Seems you reaaly got me wrong.18:19
RaimuSorry from me too.18:27
DocScrutinizerI see hundereds if not thousands of users installing your "testing .deb for brave users", no 1337 luser will miss to install sth that's supposed to make his device more 1337, esp when it's attributed "DANGER! testing only! ONLY for *brave* users"18:29
Lava_Croftyou guys want a laugh?18:30
Lava_Croftmy washing machine beat my AKG headphones18:30
DocScrutinizerand it's exactly those lusers that then go whining about "maemo is such a crap, reboots 5 times a day for me", while they conveniently ignore that they ignored the warnings all over on 70% of their installed pkgs18:30
*** Sicelo_ has quit IRC18:31
DocScrutinizernot saying we give a shit about those lusers' whining, but you won't see any useful bug reports from them, while they damage CSSU/maemo image18:31
RaimuLava_Croft: In which contest?18:31
freemangordonWell, according to my experience re 720p and SR there are more than enough "good type" people willing to test new stuff.18:32
Lava_Croftan unplanned contest18:32
Lava_Croftseems fate wanted to see if they would survive a trip through the washing machine18:32
Lava_Croftat 75celsius18:32
freemangordonAnd the others are easy to be filtered out18:32
*** Sicelo_ has joined #maemo-ssu18:33
*** nox- has joined #maemo-ssu18:33
DocScrutinizerfreemangordon: don't forget to keep an eye on dsme auto-restart counters, anyway18:34
DocScrutinizerotherwise a lot of segfaults might go unnoticed18:35
freemangordonDocScrutinizer: and lets not talk about CSSU's image, trust me, I am angry more than you on some recent developments ;)18:35
DocScrutinizerok18:36
freemangordonand not so recent18:36
freemangordonDocScrutinizer, thank for the advice re restart counters, will use that18:37
freemangordonthanks*18:37
DocScrutinizerfreemangordon: just to mention it, not saying it's mandatory or a particularly good idea: I occasionally built test versions of stuff that were intentionally missing features, had an expiration timer, or in any other way somehow made sure lusers won't keep them despite the beta state they're in18:59
DocScrutinizerhelpdesk frequently sent me thanking notes for that19:01
freemangordon:)19:06
DocScrutinizerthose users experienced enough to deal with the immanent problems of a beta are also capable of building a non-crippled version on their own. Alas usually some of them tend to publish it then19:06
freemangordonwell, I don't think there are many n900 users who are able to build new toolchain in scratchbox. And we know all of them ;)19:08
Lava_Crofti for one like beta software and dont really mind shit breaking19:10
Lava_Croftbut im not really going anywhere near compiling stuff19:10
freemangordonok, lets wait for a week (or less if I come here saying "I am stupid").19:10
DocScrutinizerI don't think investigating and double- and triple-checking stuff is stupid. If we wouldn't do that we'd not see any progress at all19:12
DocScrutinizerso maybe s/I'm stupid/I've been over-optimistic/19:13
*** Sicelo_ has quit IRC19:28
*** m0use_ has joined #maemo-ssu20:19
*** m0use has quit IRC20:19
*** andre__ has quit IRC20:33
*** andre__ has joined #maemo-ssu20:34
*** andre__ has joined #maemo-ssu20:34
*** wmarone_ has joined #maemo-ssu20:36
*** wmarone has quit IRC20:36
Lava_Croftheh, hack20:47
Lava_Crofthttp://blog.makesyouhappy.org/wp-content/uploads/2011/11/white_n900_03.jpg20:47
Sicelolooks nice. :)]20:52
Siceloi remember someone on tmo changing their case20:52
*** Kaptenen has joined #maemo-ssu21:19
*** modeless has joined #maemo-ssu21:56
arceanmerlin_1991: bugs #12085 #11974 #12445 fixed22:01
*** mase76 has quit IRC22:06
*** mase_76 has quit IRC22:06
*** mase76 has joined #maemo-ssu22:06
*** mase_76 has joined #maemo-ssu22:07
*** modeless has quit IRC22:07
*** Pali has quit IRC22:25
*** arcean has quit IRC23:13
*** modeless has joined #maemo-ssu23:47

Generated by irclog2html.py 2.15.1 by Marius Gedminas - find it at mg.pov.lt!