[01:25:50] <[7]> user890104: looks like we finally have a fully working nano4g umsboot :) [01:26:48] <[7]> I guess taht osos problem might be due to it trashing (or rather not creating) some structures in memory that the OF expects [01:26:56] <[7]> (same applies for diskflsh, which is based on the same framework) [01:27:26] <[7]> user890104: can you try to build an itunes bootstrapper from the classic v0.2 one? [01:27:47] <[7]> I'd like to give that to the next few people who complain about it not mounting [01:53:40] *** Quits: n1s (~n1s@rockbox/developer/n1s) (Quit: Ex-Chat) [02:03:19] *** Joins: clustur (~logger@c-68-53-250-91.hsd1.tn.comcast.net) [02:03:19] *** Quits: clustur (~logger@c-68-53-250-91.hsd1.tn.comcast.net) (Remote host closed the connection) [03:30:39] <[7]> user890104: do you happen to have a working emcore build for nano4g? [03:30:50] <[7]> svn head freezes with a white screen [03:53:35] <[7]> looks like an lcd driver bug + vbus detection being broken [03:59:36] <[7]> hm, looks like some of my recent changes broke umsboot on the nano2g again ;/ [04:03:00] <[7]> ok, well that was trivial [04:05:55] <[7]> https://mega.co.nz/#!0MwAGSCZ!XKSwMwJWS4wvbRk_nTms5h5JVkJajHEtAQ1338HAx-o [04:06:22] <[7]> user890104: can you build the windows bootstrappers based on that? [04:06:51] <[7]> the dfu images should be stable enough for installer production use [06:45:04] *** Quits: [7] (~quassel@rockbox/developer/TheSeven) (Disconnected by services) [06:45:17] *** Joins: TheSeven (~quassel@rockbox/developer/TheSeven) [08:03:18] *** Joins: clustur (~logger@c-68-53-250-91.hsd1.tn.comcast.net) [08:03:19] *** Quits: clustur (~logger@c-68-53-250-91.hsd1.tn.comcast.net) (Remote host closed the connection) [09:07:34] TheSeven: http://pastie.org/pastes/8195894/text?key=j147fyqupsn7bkushhkv6a [09:08:10] http://www.datafilehost.com/d/bf8bd86b [09:22:59] flashing my classic with the latest release, using umsboot 0.2 leaves it in bootrom dfu mode [09:25:13] also with r855 [09:29:54] i booted rockbox, so i can shut doown the ipod :) [09:30:43] also, flashing an installer with the old umsboot, doesn't bring it out of dfu mode [09:31:19] i guess the NOR flash is corrupted somehow, hopefully i have a backup of it somewhere [09:50:22] *** Quits: [Saint] (~saint@rockbox/user/saint) (Remote host closed the connection) [09:51:22] *** Joins: [Saint] (~saint@rockbox/user/saint) [10:13:37] *** Joins: STeeF (~STeeF@office.hostnetbv.nl) [12:23:25] user890104: that sounds like the emcoreldr image wasn't signed correctly during installer generation [12:24:07] I have a working one if you need it [12:26:36] TheSeven: i've used the build server ones [12:27:04] i'd be happy if you send me a working one, so i can revive my ipod [12:27:18] so you say the buildserver current installer version softbricks the classic? [12:27:21] lemme try that... [12:27:55] but even after flashing the release ones, it still doesn't fix it [12:28:24] ouch... that's illogical [12:28:28] and the installer succeeds? [12:28:45] yes, "Flashing..." then white screen [12:28:56] on reboot, it goes straight to dfu [12:29:09] i don't remember how i installed it on my ipod an year ago [12:29:24] i remember using a build of mine at some point [12:31:12] the installer is supposed to completely wipe the flash, except for the SYSCFG [12:31:21] if the SYSCFG would be missing it would refuse to install though [12:31:39] so after that cycle there shouldn't be anything messed up remaining [12:31:53] and we kinda know that the release one works... [12:32:50] flashing the latest buildserver one using emcore runfirmware works for me [12:33:44] old umsboot took tens of seconds to mount. hm... [12:34:22] flashing using old umsboot works as well [12:34:49] new UMSboot (via DFU) mounts MUCH faster [12:35:09] but flashing using that broke it! [12:37:53] user890104: can you send me the old umsboot dfu file that you used? [12:38:16] from the release builds? [12:40:22] hm, that one indeed has mounting trouble... much more than when booted from emcore? wtf? [12:45:03] TheSeven: i used umsboot from the build server [12:45:40] basicly, i used my .exe bootstrapper with umsboot 0.2, then loaded the installer from release [12:45:49] hm, release bootstrap + r876 buildserver installer fixes it for me [12:46:21] after trying again, and it didn't work, i ran the itunes bootstrapper again, threw a umsboot .bin from the buildserver [12:46:30] then booted the installer [12:46:58] it completed successfully, then the ipod went to softbrick [12:47:25] ah, so you were still going through the new umsboot? [12:47:31] that might explain things [12:47:53] i've had a look at the nor images: they're identical except for the hash of the emcoreldr image [12:48:07] yeah, i had it embedded in the exe, and i was too lazy to either install python, or rebuild the exe [12:48:08] I guess that is crypted with a device key during installation [12:48:25] so I figure the new umsboot might leave the hardware in a state where the crypto doesn't work [12:48:46] yeah, that would explain my installation failures [12:49:40] i can dump the nor, when i get home at about 19:00 GMT+2 [12:49:54] hm, I don't think we need that [12:50:27] or i can simply run a crypto operation from emcore [12:50:40] then recover the ipod, and run the same operation again [12:50:48] so we can compare the result? [12:51:57] probably makes most sense to just fix it, and then play around with crypto from umsboot-booted emcore [12:52:03] because that's what failed during installation [12:52:30] ok, i just wanted to be completely sure that this is the exact issue [12:54:16] and what about nano4g? you said that the bootstrapper works fine? [12:55:07] i'll give it a try on my machine, just to check if it works [12:55:55] <[Saint]> TheSeven: user890104: - You guys might get a kick out of this, a revised iLike Video: http://www.datafilehost.com/d/22c04561 [12:56:14] it should work with the itunes bootstrapper for the classic, right? i guess i'll just need to change some usb ids [12:57:01] huh? [12:57:15] <[Saint]> icons and scrollbar disabled by default, you can display additional numerical volume and/or battery data in the status bar with Settings -> Theme Settings -> Status/Scrollbar -> Battery/Volume Display -> Numeric [12:57:17] you mean the launcher tool? [12:57:24] not the dfu binary itself of course [12:57:31] yes, the windows app [12:57:35] yes, that should work [12:58:01] i'm going to use your build's dfu [12:59:32] *** Quits: [Saint] (~saint@rockbox/user/saint) (Remote host closed the connection) [12:59:57] hm... so it looks like umsboot 0.2 messes up the hw state in a way that affects the crypto and that emcore doesn't reset [13:00:11] which is kinda funny, because umsboot 0.2 touches barely anything [13:00:41] *** Joins: [Saint] (~saint@rockbox/user/saint) [13:01:13] TheSeven: any idea how this should look for nano4g? "\\\\?\\usb#vid_05ac&pid_1223#87020000000001#{b8085869-feb9-404b-8cb1-1e5c14fa8c54}\\0001" [13:01:31] i can change the usb pid, but what about the other part of the string? [13:01:38] especially that guid [13:01:59] no idea what that ID is... probably related to the installed driver [13:02:08] the serial number has to be 8720 instead of 8702 of course [13:02:24] i'd just try with the same guid and if it doesn't work check device manager for the correct one [13:02:59] ok then [13:04:34] http://www.datafilehost.com/d/8b4510cd - testers with nano4gs are welcome :) [13:04:50] well, you probably need a tester with windows + nano4g :) [13:05:00] *** Joins: n1s (~n1s@nl118-168-30.student.uu.se) [13:05:00] *** Quits: n1s (~n1s@nl118-168-30.student.uu.se) (Changing host) [13:05:00] *** Joins: n1s (~n1s@rockbox/developer/n1s) [13:05:11] yeah, makes sense [13:05:32] unless they want to run it using WINE [13:05:43] with itunes drivers. right. [13:05:48] ah ... [13:06:15] btw: i extracted the itunes drivers from the package, it is straightforward when using 7-zip [13:06:28] I have a feeling that the underlying problem should be fixed in emcore, not umsboot [13:06:41] we can write instructions, so people can extract them and install them, without messing with apple's installers at all [13:06:57] we could also just fix our own drivers :) [13:07:00] even a 7-zip binary with a script or something [13:07:09] well, we need to sign them [13:07:13] nah [13:07:17] winusb [13:07:21] or use libusb/winusb signed binaries? [13:08:06] if we ignore winxp users that have never installed a wdf based driver, we can even remove those coinstallers from the whole mess [13:08:15] someone needs to properly rewrite that inf file [13:08:22] the binaries are all from microsoft [13:08:35] and we already have a bootstrapper tool that can talk to them [13:22:50] user890104: I've found a missing cache flush... [13:23:12] hm, or actually it shouldn't matter if nothing else misbehaves [13:29:18] working emcore: sha1(0x08339430, 0x00000E10) = 179E5199F090E97C0154DD0F7EACA6C4CAA6FECB [13:30:51] broken emcore: sha1(0x08217AC0, 0x00000E10) = 0000000000000000000000000000000000000000 [13:30:57] I'd say it's clear who the culprit is [13:31:52] no DMA involved [13:32:03] clock gates seem to be taken care of - IF we have caught the right one [13:33:11] disabling that gate clearly screws up the sha1 unit, so we're not totally wrong there [13:35:41] wild guess: it could be the page tables [13:54:18] *** Joins: staodd (~Steele@ti0054a380-dhcp2737.bb.online.no) [13:55:31] *** Parts: staodd (~Steele@ti0054a380-dhcp2737.bb.online.no) () [14:01:01] ok, that's it [14:01:24] not configuring the MMU => works [14:01:52] as emcore reconfigures the MMU once again, but doesn't fully regenerate the page tables, it must be the page tables that are the problem here [14:01:55] let me take a look at them [14:03:21] *** Joins: clustur (~logger@c-68-53-250-91.hsd1.tn.comcast.net) [14:03:21] *** Quits: clustur (~logger@c-68-53-250-91.hsd1.tn.comcast.net) (Remote host closed the connection) [14:04:05] exactly one byte of difference! [14:06:18] great, can you send me the build, so i can embed it in the loader? [14:06:47] this should fix the installer, right? [14:22:03] the mapping for 0x38000000-0x380fffff is marked as cacheable, bufferable [14:22:08] I'm wondering why... [14:23:25] I basically copy&pasted that code from emcore loader... [14:30:42] i see [14:30:54] ah, I think I know what happened [14:31:10] it's in fact an umsboot bug [14:34:07] Farthen: I'm afraid one of my next commits will break the build server [15:06:09] user890104: fixed version: https://mega.co.nz/#!NAoChaZA!LvrFIyi-cyTiP40U6Du7ITJpvUEITxhH5VCO2sIyKn4 [15:06:18] please rebuild the bootstrappers once again [15:06:30] TheSeven: both of them? [15:06:36] ideally yes [15:06:40] ok [15:06:40] they have the same bug [15:10:31] https://mega.co.nz/#!ko4G2JhZ!SqE5L3GxP-dCtzyoa-tAsS-aFfZ2tOkLASXfGTxaDKw [15:14:17] ah, seems like my code was spamming the i2c bus, polling for rtc data :) [15:18:44] ~100 transactions per second, yes :) [15:19:03] in fact that's what was limiting the frames per second on nano2g :) [15:21:17] hm, build server broke badly [15:21:30] the whole umsboot column of builds, also for older revisions, is gone [15:21:41] and the old toolchain doesn't seem to like the new code [15:29:07] well, umsboot column disappearing is not a bug, it's a feature :) [15:29:14] let me fix it [15:29:23] (if my ftp account still works) [15:32:05] user890104: do you have a clue why my attempt at a build server fix didn't work? [15:35:50] TheSeven: what did you change? [15:37:58] i'm trying to make the script find those new directories [15:38:28] and by script, i mean the php page, not the post-commit hook [15:38:38] because i don't have access to the latter [15:38:48] some paths have changed [15:39:05] the target build dirs are now build/umsboot//release [15:39:15] instead of just build/ [15:40:00] yeah, i'm trying to find where do i need to cut that extra part (if it exists) [15:40:19] I've put this into the makefile: [15:40:19] FLTO ?= -flto -flto-partition=none [15:40:19] ifneq ($(shell $(CROSS)gcc -flto 2>&1 | grep -- -flto),) [15:40:19] FLTO := [15:40:19] endif [15:40:30] however it doesn't seem to help [15:40:47] the build log contains a bunch of [15:40:47] cc1: error: unrecognized command line option "-flto" [15:40:47] cc1: error: unrecognized command line option "-flto-partition=none" [15:43:25] over here it does work, but not on the build server [15:53:48] well, i have no idea about gcc options [15:54:13] i just fixed the php script [15:54:47] so it will show new builds when they're available in the new dir structure [15:56:13] TheSeven: you are grepping stdout, but gcc prints its version on stderr :) [15:56:53] ah, you already tried that [15:59:33] "gcc -v --help 2>&1 |grep -- -flto" does not find any lines on my machine [15:59:49] if i change flto with fprofile (for example), it works [16:00:20] i mean, it matches some of the lines [16:03:39] woah. now that's some errors: http://builds.freemyipod.org/emcore-installer/target/ipodclassic/emcore-installer-ipodclassic-r887.log [16:04:01] user890104: that's because your compiler probably doesn't support it [16:04:20] that problem is fixed now... but we have other trouble [16:04:35] the toolchain on the build server is somehow a bit weird [16:11:00] hm it's getting better [16:37:55] user890104: I'm giving up for now. Those build server errors are just illogical. [16:38:22] there's not much point in trying to debug that without shell access to the server [16:39:15] TheSeven: ok, i'll test those new bootstrappers when i get home [17:01:55] *** Joins: ten10 (484a3b66@gateway/web/freenode/ip.72.74.59.102) [17:05:04] *** Quits: ten10 (484a3b66@gateway/web/freenode/ip.72.74.59.102) (Client Quit) [19:16:49] TheSeven: well, if you fix it… :D [19:17:45] it's acting all weird [19:17:54] I'd probably just swap out the toolchain for a newer one [19:19:54] well yeah [19:20:25] https://github.com/EliasOenal/TNT/blob/master/ [19:21:07] the gcc download is broken, so download http://releases.linaro.org/13.06/components/toolchain/gcc-linaro/4.8/gcc-linaro-4.8-2013.06-1.tar.xz first before running the script [19:48:43] *** Quits: STeeF (~STeeF@office.hostnetbv.nl) (Remote host closed the connection) [20:03:22] *** Joins: clustur (~logger@c-68-53-250-91.hsd1.tn.comcast.net) [20:03:22] *** Quits: clustur (~logger@c-68-53-250-91.hsd1.tn.comcast.net) (Remote host closed the connection) [20:18:55] TheSeven: my nano4g doesn't like your dfu image, it just reboots [20:20:32] there's a box saying UMSboot launched successfully, but that's all - the ipod is already rebooted [20:49:24] that's odd [20:51:20] i can try reuploading with ipoddfu [20:54:05] * user890104 says hello to his ubuntu VM [21:47:42] user890104: would you be interested in co-developing a userspace USB storage solution for emcore? [21:47:59] that allows you to either mount the ipod itself, or a disk (floppy, cdrom, hard disk) image stored on it? [21:49:55] that would be something that these devices could actually be good for these days [21:51:07] TheSeven: sounds interesting, i don't know if i can help much with that, but i'll definetely give it a try [21:51:35] so we need to add usb api to emcore's syscalls first, right? [21:51:58] yes, I'll try to take care of that (and at the same time free up the debugger endpoints) [21:52:04] and then write an app that uses it, and emulates different kinds of usb storage devices [21:52:15] exactly, that's where you could help [21:52:27] i see [21:52:38] especially since you have already implemented a filechooser with libui etc. [21:53:50] ah, that one? right, i almost forgot about it [21:54:22] did i add it to the svn? [21:54:49] no idea [21:55:08] it should be somewhere in my VM's working copy then [21:55:36] so yes, i can help you with the app [21:56:01] do you have a rough idea how usb mass storage works? [21:56:30] http://websvn.freemyipod.org/filedetails.php?repname=freemyipod&path=%2Fumsboot%2Fsrc%2Fapp%2Fumsboot%2Fums.c [21:56:41] that is roughly what will need to be implemented on the app side of it [21:57:47] and http://websvn.freemyipod.org/filedetails.php?repname=freemyipod&path=%2Fumsboot%2Fsrc%2Fapp%2Fumsboot%2Fusbglue.c [22:01:17] ah, so it's basicly sending scsi commands over usb? [22:01:21] yes [22:02:00] i remember dealing with these when testing ipodscsi [22:03:09] http://www.usb.org/developers/devclass_docs/usbmassbulk_10.pdf [22:03:10] btw: can you test the linux version of ipodscsi, and possibly merge it with yours? [22:03:44] some guy wrote it at some point, and told us that we can include it in our code if we wish [22:05:55] v1: https://dl.dropboxusercontent.com/u/26417574/ipodscsi_linux/main.c [22:05:58] v2: https://dl.dropboxusercontent.com/u/26417574/ipodscsi_linux/main-v2.c [22:07:27] i can also do that with some #ifdefs, if there isn't a better way to merge them [22:08:16] hm ideally one would try to abstract the scsi accesses into some functions [22:11:14] can you test v2? i'm having issues attaching usb devices to virtualbox