#qi-hardware IRC log for Wednesday, 2013-04-10

wpwraklet's see what happens if i set kseg0 to write-through ...08:13
wpwrakkernel gets a little slower. good so far. now let's try IO ...08:15
wpwrakhmm, 10% out of 50% possible speedup. kinda disappointing :-(08:17
wpwrakcrazy measures should have a more satisfying return08:18
hellekin Vote for wpwrak :)09:58
whitequarkgrrrr09:59
whitequarkcanada post can't seem to distinguish russia and uk09:59
larscI can understand when they mix up austria and australia, but how do you get the uk and russia confused?10:38
whitequarkhttp://www.canadapost.ca/cpotools/apps/track/personal/findByTrackNumber?execution=e1s1 30459917676510:39
whitequarkI have *no idea*10:39
viricwhitequark: uk for ukraine, or for united kingdom?10:58
whitequarkviric: united kingdom10:59
viricUK = United Kingdom = Royaume-Uni = RU = Russia10:59
viriccould be that?11:00
larsc...11:00
viricIt's very usual to shorten Royaume-Uni to RU, (same in Catalan; Regne Unit = RU)11:01
viricand then you have UK for Ukraine too11:02
viricnot that difficult to mix up :)11:02
whitequark... indeed.11:03
larscdid you write RU in your address field?11:03
whitequarklarsc: it was "Russia".11:04
viricRussia is harder to mix up11:04
whitequarknot impossible evidently :/11:04
Action: whitequark has run sloccount on gcc sources11:05
larscbut at least they can't say its your fault11:05
whitequark2848901 gcc             ansic=1727123,ada=738502,cpp=247495,f90=85207,exp=16009, objc=14691,fortran=9895,ml=3007,sh=2446,awk=1880,pascal=1005,perl=913,asm=416,lex=200,haskell=11211:05
whitequarklarsc: indeed.11:06
whitequarkomg, the C lexer is... really really simple11:08
viricthat put C in a quite safe position, among programming languages11:09
whitequarkalso, I wonder if anyone could explain the rationale of this:11:09
whitequark /* Print GENERIC declaration (functions, variables, types) trees coming from the C and C++ front-ends as well as macros in Ada syntax.11:09
whitequarkin the gcc-4.8.0/gcc/c-family/c-ada-spec.c.11:10
whitequark... utf-32 string literals?11:11
whitequarkok let's not stare into the abyss any longer.11:11
larscbetter quite before it stares back into you?11:15
whitequarklarsc: yes.11:16
Action: whitequark shudders at the very thought of 7.3MLOC of GCC staring at him.11:16
larsc:)11:17
whitequarkthere is something Lovecraftian about that. an ancient beast waking up from its eternal dream, only to cast a single stare and drive an innocent programmer on the verge of insanity11:17
wpwraksee, better hurry and lose your innocence. then you're safe :)12:11
whitequarkI wonder what would constitute losing innocence in the context of GCC...12:16
whitequark(also https://pp.vk.me/c410422/v410422132/82c0/Yt400Y9QKN4.jpg)12:16
wpwrakmarvelous !12:18
whitequarkomg.14:16
whitequarka polite postal worker (!) who, having seen an attached paper which stated a weight difference (0.9kg vs 0.6kg) asked me to check the contents and said she'll help with paperwork if something's actually was missing14:18
whitequark(!!)14:18
pcercueithere are some nice people out there :)14:19
wpwraks/polite/fake/ ? :)14:19
whitequarkwpwrak: haha14:19
whitequarkalso they did not lost the parcel despite me writing the wrong zipcode14:20
larscwhitequark: you'll be on one of these hidden camera shows14:20
whitequarkand, in a previous encounter with EMS (a subsidiary of the russian post), when I've had two parcels to receive and forgot to write the flat # on one of them, a courier who delivered the second one has remembered it and offered to redeliver14:21
whitequarkI wonder if they're they still a government-sponsored organization?14:22
wpwraklarsc: he should go easy on these mind-altering drugs, particularly the strong ones14:22
whitequarkhuh, 2/3 of connections are suddenly ECONNREFUSED. not sure if just crappy connection or malfunctioning censorship engine15:22
whitequarkregardless, openvpn solves this :)15:22
viricmeh16:32
viricUBI error: ubi_read_volume_table: the layout volume was not found16:32
viricoh the new qemu added a NAND to the default mips board16:39
viricCreating 3 MTD partitions on "physmap-flash.0":16:40
viric0x000000000000-0x000000100000 : "YAMON"16:40
viric0x000000100000-0x0000003e0000 : "User FS"16:40
viric0x0000003e0000-0x000000400000 : "Board Config"16:40
kyakwpwrak: spi-gpio-atben and spi_atben_gpio - i'm already lost :)17:24
kyakbtw is there a naming convention for kernel modules at all?17:25
larscfor some17:31
larsce.g. some subsystems use a common prefix17:32
larsclike i2c-foobar17:32
wpwrakkyak: yeah, it's a little messy :) also, spi_atben is somewhat correct because it implements a SPI driver, but then spi_atben_gpio only puts things together17:37
wpwrakkyak: but i think spi_atben can die now. it was the very first one i wrote.17:38
wpwrakand spi_atben_gpio needs a less confusing name17:38
wpwrakspi-* are all real SPI drivers17:38
kyakis there a chance atben and atusb drivers will be accepted by upstream?17:41
kyakis it even something you want to do, or we don't really care?17:42
wpwrakoh, i want them upstream. i don't expect undue difficulties. just need a bit more testing.17:45
wpwrakand atben has a bunch of at86rf230.c fixes it semi-depends on as well (as in "i already know it'll crash and burn before long if they're not there"). they sould probably go in first.17:46
kyakcool :)17:47
wpwrakof course, atben has the added issue that some things are missing in mainline for proper ben support. i do my development on a board without lcd, so i just applied the two patches i need to get at least that far. but that kernel probably has no lcd. and it also doesn't have USB.17:52
wpwrak(i flash directly via usb-boot, which is very easy if you have idbg)17:53
wpwrak(and can of course be the mother of all nightmares if you don't :)17:53
kyakdo you mean that your driver might affect USB?17:56
wpwrakno, but the net-next kernel doesn't have the patch for the jz4740's USB device. so anything you build from that source will be a somewhat limited ben.17:59
wpwraknot sure where that driver is in the pipeline, if it's on the way into mainline at all17:59
kyakok, now i understood :)18:00
kyakthough i thought that most of the patches went upstream18:01
kyakexcept for partitions maybe18:01
wpwrakmaybe it's somewhere in a tree that hasn't been merged yet. we have quite a number of patches in the qi-gw owrt repository that aren't in mainline. e.g., several NAND performance improvements.18:26
pcercueihmm18:54
pcercueihow can I stop a blocking read() call from a different thread?18:55
larsckill the thread19:55
larscor send any other signal19:55
pcercueiwell, I tried that :)19:55
pcercueipthread_kill(thd, SIGTERM)19:55
pcercueidoesn't stop the read() call :(19:55
larscit should, at least if the signal is not blocked19:55
pcercueiwell, SIGTERM should never be blocked, no?19:55
larscSIGKILL is unblockable19:55
larscSIGTERM is just like any other signal19:55
pcercueihmmm19:55
pcercueiok19:55
larscit's also possible that your backend does a non-interruptible sleep19:56
pcercueiso I believe I just have to use sigaction()?19:56
larscprobably19:56
larscor you just use poll and friends19:57
pcercueipolling would be a bit... dirty19:58
larscno, poll()19:59
larsca.k.a select()19:59
pcercueilarsc: sigaction() does affect the whole process, not only a thread20:06
viricpcercuei: I think you can close the fd20:07
pcercueiviric: oh, good idea20:08
pcercueithat would probably work20:08
pcercueiviric: doesn't work :(20:27
pcercueicalling close() on the file descriptor doesn't stop the read()20:28
larscwhat are you reading on?20:29
pcercueia file descriptor obtained with inotify_init()20:30
virichm ok20:30
viricthen I'm with lars20:30
viricselect or signal20:30
larscaccording to the code inotify_read is interruptible20:32
pcercueithat's why I don't understand why I can't interrupt it ;)20:39
viricinterruptible means that on a signal it will unblock, no?20:40
viricmaybe you have set SA_RESTART for SIGTERM ?20:40
pcercueiwell, my code doesn't set it20:41
viricok20:41
viricI never understood how threads and signals cope together though20:43
larscwpwrak: be careful if you ever want to buy a poodle 21:14
larschttp://news.yahoo.com/blogs/sideshow/ferret-poodle-argentina-125901615.html21:14
wpwrakbwahaa. of course, he deserves it for buying at La Salada, the XXL market for fences (german "Hehler") and scammers.21:23
--- Thu Apr 11 201300:00

Generated by irclog2html.py 2.9.2 by Marius Gedminas - find it at mg.pov.lt!