**** BEGIN LOGGING AT Thu Sep 08 02:59:56 2005 Sep 08 03:31:38 is there an easy way to install openssl-dev in the nslu2 "mastermakefile" env ? Sep 08 03:34:46 mr_claus: Mind making a patch for icecast? Sep 08 04:37:27 cool Sep 08 04:37:34 * NAiL set up basic hotplugging Sep 08 04:37:45 for the devices that I own, atleast ;) Sep 08 04:38:21 Now my IXUS is automatically emptied whenever I connect it to the slug Sep 08 04:39:19 auto-format? :D Sep 08 04:40:18 haha Sep 08 04:40:36 It moves all the pictures from the ixus and puts them on a share Sep 08 04:40:46 u dont happend to know wtf i can find libcrypto do you? :) Sep 08 04:41:06 dl it precompiled somewhere maybe? Sep 08 04:41:07 ipkgfind.nslu2-linux.org is a nice place to start looking for files ;) Sep 08 04:41:15 i dont got a nslu2 Sep 08 04:41:20 i have to crosscompile Sep 08 04:42:10 donno if the mastermakefile-thingie can use ipkg's etc!? Sep 08 04:43:04 You want to crosscompile libcrypto? For what target? Sep 08 04:43:32 the master makefile *makes* ipkgs Sep 08 04:43:45 nono, i need libcrypto to crosscompile an app Sep 08 04:44:14 (its a pain to crosscompile openssl manually) Sep 08 04:45:26 "bb openssl" should crosscompile openssl, and put the libraries in staging Sep 08 04:48:13 ah Sep 08 04:48:15 superb Sep 08 04:48:20 thanks alot :) Sep 08 04:49:32 hm, in a special dir, or? Sep 08 04:49:54 cd openslug && source setup-env && bb openssl Sep 08 04:50:08 super :)) Sep 08 05:11:13 03hrw 07org.openembedded.dev * r720e5c15... 10/packages/meta/ (opie-collections.inc opie-image.bb): Sep 08 05:11:13 opie-image: added opie-freetype into OPIE_BASE Sep 08 05:11:13 - was missing due to recent change in opie-freetype and opie-ttf-support depends Sep 08 05:11:54 re Sep 08 05:13:16 NAiL: oh, you're maintaining icecast? Sep 08 05:13:21 NAiL: it seems Sep 08 05:13:50 NAiL: you would do a patch with diff or with sed? Sep 08 05:14:19 grr, xinetd.org is down Sep 08 06:16:17 03mickeyl 07org.openembedded.dev * r50b29d3f... 10/packages/efl/ (5 files): efl: *) September is the 9th month... *) ship emotion plugins Sep 08 06:20:40 What is the ETA on openslug 2.6 (source) Sep 08 06:32:17 VoodooZ_Log: You mean OpenSlug $NEXTRELEASE-beta? Sep 08 06:33:31 mr_claus: Looks like one has to hack configure.in, I guess. sed is considered a 'last resort' in .bb's Sep 08 06:34:24 i compiled 2.6beta with udevd, with 2.3 it worked without any problemns now the daemon will not be started Sep 08 06:34:52 Yes, I guess it's time to revisit that issue too. Now I have a working devslug again. Sep 08 06:35:35 hotplug should start udevd at startup i think Sep 08 06:36:09 You've read the stuff in the udev bb? Sep 08 06:36:38 not yet because it worked with 2.3 so i thought there is no reason why it should not run with 2.6 Sep 08 06:37:14 * mr_claus is reading the bb Sep 08 06:37:46 The FIXME's should give you a hint IIRC Sep 08 06:39:05 ok, so i have to start hotplug and udev with the init script? Sep 08 06:39:56 #FIXME d) eventually hotplug should no longer be used at all, all agents shall be converted to udev rules Sep 08 06:40:39 ok, at the moment there is no hotplug running Sep 08 06:40:53 i will try to start udevd with the init script only Sep 08 06:41:01 VoodooZ_Log: no releasedate is set. Now that the topic of udev has come up, I guess it's time to try to squeeze it in before a release Sep 08 06:41:50 Then test and eventually release. I'm not the one who decides that though ;) Sep 08 06:43:08 mr_claus: Don't know how the initscript is supposed to be... Something about that "netlink" stuff... Sep 08 06:56:24 hmm, there is not much free space free on the image Sep 08 06:57:41 start off by removing the kernel ;) Sep 08 06:57:51 :) Sep 08 06:57:55 good idea :) Sep 08 06:58:41 <[g2]> unless you're running APEX then you can can pickup 1.25 MB from the old kernel and boot directly to the kernel in the jffs2 partition :) Sep 08 07:03:10 i could remove reiserfs too Sep 08 07:03:37 <[g2]> nod on the reisefes Sep 08 07:05:58 there's a bit to save by fixing the fsck.ext2/3 files in /sbin too Sep 08 07:06:02 97 -rwxr-xr-x 1 root root 99172 Jan 1 1970 mke2fs Sep 08 07:06:03 97 -rwxr-xr-x 1 root root 99172 Jan 1 1970 mkfs.ext2 Sep 08 07:06:03 97 -rwxr-xr-x 1 root root 99172 Jan 1 1970 mkfs.ext3 Sep 08 07:06:15 monotone merge -b org.openembedded.dev => monotone: trying 3-way merge => monotone: misuse: no unique private key for cert construction;( Sep 08 07:06:16 mkfs, not fsck :P Sep 08 07:06:56 well... fsck ext2/3 is duplicated too, I see... Sep 08 07:11:24 By making sbin/fsck.ext3 -> fsck.ext2 and mkfs.ext[23] -> mke2fs, you save ~280 Sep 08 07:11:27 k Sep 08 07:13:20 Tuborg: you need a private key to merge... Look at the wiki in the development section Sep 08 07:14:13 NAiL, Thanks. I'm fed up of my builds breaking so I'll be using the source release instead. Sep 08 07:14:33 Ah, ok Sep 08 07:14:55 VoodooZ_Log: If you can help us get udev running properly, it'll speed it up ;) Sep 08 07:15:19 whoa... removing reiserfs support saves like a mb of flash Sep 08 07:15:23 the default install worked ok for my app. Sep 08 07:15:31 I know. that's why I always remove it! Sep 08 07:15:33 I know, but should I resolv the merge selv? Sep 08 07:16:08 Tuborg: most of the time it'll figure it out by itself Sep 08 07:16:20 reiserfs should be an option if you ask me. This thing is getting bloated like MS-Winblows~! Sep 08 07:16:37 Had no idea it was that large... Sep 08 07:17:34 Yay! Sep 08 07:22:26 * NAiL suddenly has 2012k free Sep 08 07:24:45 2020k without portmap Sep 08 07:25:58 i think reiserfs could be installed seperately, there is no need to keep it in the basic system Sep 08 07:26:20 because most would be use ext3 out of the box i think Sep 08 07:26:57 especially on a non-x86 system Sep 08 07:26:58 but if the external device should be running and turned with reiserfs then it would be needed Sep 08 07:28:38 VoodooZ_Log: is a /sbin/udevsend before exec /sbin/init enough to use udev? Sep 08 07:28:49 personally, anyone using reiser on a 266MHz ARM needs their head testing Sep 08 07:28:59 with all due respect to those people :) Sep 08 07:29:23 then ntfs would not be needed too Sep 08 07:29:23 Use ext3 with the -O dir_index option Sep 08 07:29:28 could be an optional package Sep 08 07:29:50 <[g2]> blaster8 why the issue with reiserfs on ARM ? Sep 08 07:30:01 none specific Sep 08 07:30:14 but it is too CPU intensive for such a low power computer Sep 08 07:30:18 <[g2]> I think one only needs to change the default _EXTRAs Sep 08 07:30:30 and also it has only been heavily tested on x86 Sep 08 07:30:36 <[g2]> blaster8 you've measured reiser3 and reiser4 ? Sep 08 07:31:16 how many ppl will be using their Slugs for running gentoo portage emerge sync or high end databasing application where fs performance becomes an issue Sep 08 07:31:46 given that the default slug can hardly saturate its Ethernet connection, let alone its USB Hard drive connection Sep 08 07:31:48 ok, now udev is working, where are the files in /dev from? i have to delete some of them in the buildsystem Sep 08 07:33:43 <[g2]> blaster8 the Slug can easily saturate and Ethernet connect, disk I/O on the slug is an issue Sep 08 07:34:12 <[g2]> However, there are other devices out there that will run nearly the identical image Sep 08 07:34:31 <[g2]> The Linksys RV rouuter series has IXP425s at 266 and 533 Sep 08 07:34:36 mr_claus: ./files/device_table-openslug.txt Sep 08 07:34:39 <[g2]> I'm building a board right now Sep 08 07:35:38 are these systems running with the openslug image too? Sep 08 07:35:42 NAiL: thx Sep 08 07:36:41 [g2]: yeah, but it doesn't hurt having it as an option.. Sep 08 07:36:50 considering it takes 1mb flash Sep 08 07:37:31 perhaps not the same config file (openslug.conf) could be used, openslug is nslu2 it seems and nslu2 is slow Sep 08 07:38:04 <[g2]> mr_claus, OpenSlug is a "branded" nslu2 release Sep 08 07:38:47 <[g2]> However, the OpenSlug rootfs would runs just fine on these kind of devices with just a kernel port and possibly some drivers Sep 08 07:39:15 <[g2]> I'd that back in March with an OEM board and it only took a couple hours for the port Sep 08 07:39:16 why not fork openslug.conf if other devices needs reiserfs? Sep 08 07:39:59 <[g2]> mr_claus simply changing the default _EXTRA in the conf file causes it not to be built if you are building OpenSlug Sep 08 07:40:21 <[g2]> I haven't done that lately, but it's totally trivial to do Sep 08 07:41:16 i know, i have done that, but the real question in this discussion is "need the nslu2 the reiserfs integrated or optional" Sep 08 07:42:08 i think we're only talking what should be the default, with or without reiser Sep 08 07:43:26 and i think, most people will not need reiserfs so it's better to take it out by default, but that's only my opinion, i have no problem with taking it out by hand Sep 08 07:45:04 NAiL: do you think i can take out all devices of the device table file? will udevd handle all needed device files? Sep 08 07:45:28 <[g2]> mr_claus My take on the situation was the jbowler put it in because he used it and it is trivial to take out. Sep 08 07:46:00 <[g2]> given that he created a streamlined uCSlugC release, it's probably not used that much by him anymore Sep 08 07:46:29 mr_claus: I think you need *atleast* two devices by default... console and null. Try ;) Sep 08 07:46:30 03mickeyl 07org.openembedded.dev * rfbd660b2... 10/packages/efl++/ (5 files): efl++: bump to v0.0.5 Sep 08 07:46:38 <[g2]> also, given that the jffs2 rootfs is very full, it would be an excellent candidate in my mind to be made optional again Sep 08 07:46:55 yeah Sep 08 07:46:59 [g2]: ok, thx Sep 08 07:47:04 NAiL: i will try that Sep 08 07:47:18 I'm going to fix e2progs to make symlinks... Sep 08 07:51:06 "monotone genkey Tuborg@home.no" Then "monotone merge -b org.openembedded.dev -k Tuborg@home.no" worked for the "monotone: misuse: no unique private key for cert construction" Sep 08 07:54:45 Bet I can commit bugs now,lol. Sep 08 07:55:35 You'd need write access to the oe repo Sep 08 08:04:07 package ipkg-native-0.99.152: completed;) Sep 08 08:09:32 mr_claus, no idea. I just installed the package. Sep 08 08:12:12 I fully agree with blaster8 about the reiserfs. It should be made optional. Like the rest. Sep 08 08:18:43 wtf? e2fsprogs defines sbindir as /sbin... Sep 08 08:19:49 Shouldn't the openslug source tarball work out-of-the-box? Sep 08 08:20:05 yeah, doesn't it? Sep 08 08:20:22 well, I must be stupid but it doesn Sep 08 08:21:08 where does it fail? Sep 08 08:21:36 It gives me this for every package: ERROR: [Errno 2] No such file or directory: '/local/openslug_2.5/openembedded/packages/zlib/*.bb' while parsing /local/openslug_2.5/openembedded/packages/zlib/*.bb Sep 08 08:23:10 I'll check the wiki for info. I kind of rushed it in. Sep 08 08:23:36 Stuffed BBPATH or something, I guess... Sep 08 08:25:05 I guess I was expecting it to work with a simple untar and make but guess not. Sep 08 08:34:25 if i delete tmp/work/initscripts....., should it not to be recreated with "make build-openslug"? Sep 08 08:34:51 delete tmp/stamps/initscripts* too Sep 08 08:36:56 ok, thx i missed that Sep 08 08:41:13 03koen 07org.openembedded.dev * r4c3dd1fc... 10/packages/gaim/gaim_2.0.0-cvs-mco1.bb: Sep 08 08:41:13 gaim 2.0: update bb to fetch patches out of hh.org cvs via cvsweb, courtesy Mads Olesen. Sep 08 08:41:13 closes bug #310 Sep 08 09:01:09 03nail 07org.openembedded.dev * r2ed62cd8... 10/packages/e2fsprogs/e2fsprogs_1.38.bb: Sep 08 09:01:09 e2fsprogs: Fix the use of sbindir (changed to base_sbindir) Sep 08 09:01:09 -mke2fs: Change mkfs.ext[23] to symlinks Sep 08 09:01:09 -e2fsck: Chanke fsck.ext[23] to symlinks Sep 08 09:16:06 03nail 07org.openembedded.dev * rca142ce9... 10/conf/distro/openslug.conf: Sep 08 09:16:06 openslug.conf: Remove reiserfs support from default image. Sep 08 09:16:06 Saves ~1mb in flash. If anyone wants it, they can ipkg install it. Sep 08 09:20:22 mr_claus: how's the udev stuff coming along? Sep 08 09:21:22 hmm, what to use the extra 1MB for? Sep 08 09:24:35 NAiL: i have only these two nodes in the file but the image contains all others too, i think i have to delete a package in work and stamp Sep 08 09:26:08 mr_claus: did you remove tmp/rootfs first? Sep 08 09:26:51 NAiL: no, thats the mistake, thx Sep 08 09:51:04 03koen 07org.openembedded.dev * r9fe75450... 10/conf/distro/familiar-0.8.3.conf: conf/distro/familiar-0.8.3.conf: add missing ", thanks to Matt Reimer for spotting it Sep 08 09:52:48 NAiL: deleted rootfs but same result Sep 08 09:53:15 hmm Sep 08 09:53:18 then I don't know Sep 08 09:53:35 i will try to check it out but i have to leave soon and will be back later Sep 08 10:05:30 bbl Sep 08 11:21:10 03jbowler 07org.openembedded.dev * rd69e4956... 10/packages/uclibc/ (4 files in 2 dirs): Sep 08 11:21:10 uclibc: put ARM thumb patches from 0.9.27 into 0.9.28 Sep 08 11:21:10 Some of the thumb patches had already been pushed into the uclibc release, Sep 08 11:21:10 this commit adds corrected/updated versions of most of the ones which had Sep 08 11:21:10 not been accepted. The controversial 'change the ldso symbol resolver' Sep 08 11:21:11 patch is not in this set - it will be re-worked to match the changes being Sep 08 11:21:13 made in binutils for the same problem. Sep 08 12:05:13 ka6sox-office: is there a command in ppcboot to activate the NIC? Sep 08 12:43:07 yes but its specific to the hardware that you are using...SCC1 or SCC2 for example Sep 08 12:43:47 ka6sox-office, g+, Marvel Yukon Sep 08 12:44:17 the kernel sets up the Yukon not ppcboot Sep 08 12:44:50 ppcboot usually deals with internal MAC's like SCC1 or SCC2 on the Proc chip Sep 08 12:45:08 I"m not sure if it knows about the yukon Sep 08 12:45:16 I'll have to investigate when ours arrive. Sep 08 12:45:23 ok, so no change that it supports the yukon then?!? Sep 08 12:45:36 but i could upload a kernel image using kermit... Sep 08 12:45:52 via serial? Sep 08 12:45:57 or tftp? Sep 08 12:45:59 yes Sep 08 12:46:02 serial Sep 08 12:46:11 03koen 07org.openembedded.dev * rcf261b9a... 10/packages/galago/ (libgalago-0.3.3/no-check.patch libgalago_0.3.3.bb): libgalago: add libgalago to OE Sep 08 12:46:13 because for tftp the NIC must first work... Sep 08 12:46:33 the ethernet may not be enabled yet..I need to look at the ppcboot stuff. Sep 08 12:56:11 03koen 07org.openembedded.dev * rf4f18562... 10/packages/galago/ (libgalago-0.3.3/no-check.patch files/no-check.patch): galaga no-check.patch: move to files, since it can be shared with galago-daemon Sep 08 13:06:11 03jbowler 07org.openembedded.dev * rb8529852... 10/packages/freeze/files/freeze: Sep 08 13:06:11 freeze: Add support for uclibc_svn build Sep 08 13:06:11 freeze did not have patterns to handle svn, as opposed to cvs, packages. Sep 08 13:06:11 This commit just adds the pattern required to recognise the uclibc_svn Sep 08 13:06:11 package. Sep 08 13:20:06 VoodooZ_Log: ping? Sep 08 13:36:16 03koen 07org.openembedded.dev * r45599c4d... 10/packages/galago/galago-daemon_0.3.4.bb: galago-daeom: add 0.3.4 to repo Sep 08 13:36:19 03koen 07org.openembedded.dev * r5d7867f5... 10/packages/galago/libgalago_0.3.3.bb: packages/galago/libgalago_0.3.3.bb: fix staging and inherit pkgconfig Sep 08 14:55:15 03mcdmx * 10unslung/make/ds101-bootstrap.mk: ds101-bootstrap: toggle bs and skip args to speed up file extraction Sep 08 15:03:48 03orepvik * 10unslung/make/ipkg.mk: Fix ipkg-clean (Forgotten to replace with ipkg) Sep 08 15:08:26 03orepvik * 10unslung/make/ds101-bootstrap.mk: Make ds101-bootstrap handle bumps in revision Sep 08 15:22:27 03sjmelhuish * 10unslung/make/oww.mk: Bumped to latest oww version (0.81.7) Sep 08 15:56:12 03florian 07org.openembedded.dev * r3d296e3e... 10/packages/gpe-login/ (files/busybox-bad-perms.patch gpe-login_0.82.bb): gpe-login: Use patch by Philippe De Swert to work around bugs in busybox related to handling of links. Sep 08 17:32:34 So, we need someone to generate the R29 to R63 linksys kernel source patch ... Sep 08 18:46:08 03jbowler 07org.openembedded.dev * r4b93f0ae... 10/packages/uclibc/uclibc-0.9.28/thumb-resolve.patch: (log message trimmed) Sep 08 18:46:08 uclibc: fix STT_ARM_TFUNC handling in the uclibc ld.so Sep 08 18:46:08 STT_ARM_TFUNC handling was broken in uclibc because it is an ARM specific Sep 08 18:46:08 hack and uclibc code did not have support for it. Added the support in a Sep 08 18:46:08 way which I hope will be acceptable to uclibc and which should remain Sep 08 18:46:11 compatible with forthcoming binutils changes. ARM Thumb code is fully Sep 08 18:46:13 working with this commit with the possible exception of uclibc/gcc itself Sep 08 18:56:13 03jbowler 07org.openembedded.dev * r67f2ba05... 10/packages/uclibc/ (3 files in 3 dirs): Sep 08 18:56:13 uclibc_svn: fix the SVN version to build the uclibc SVN head Sep 08 18:56:13 The SVN version now copies 0.9.28 patches, the armeb case should work, Sep 08 18:56:13 other architectures almost certainly require the uclibc-cvs/*/uClibc.config Sep 08 18:56:13 copy of the config file to be deleted (if this is done the build will pick Sep 08 18:56:14 up the config file in uclibc-0.9.28). Sep 08 20:01:04 03nail 07org.openembedded.dev * re55a199a... 10/conf/distro/openslug.conf: disapproval of revision ca142ce9ae4b3c4535d5ed6d23aed2559ef91c6b Sep 08 22:28:07 03rpedde * 10unslung/ (make/mt-daapd.mk sources/mt-daapd/itunes5.patch): Fixes for iTunes 5 Sep 08 22:31:44 rpedde, excellent! Sep 08 22:34:56 nice Sep 08 22:37:58 day late and a dollar short, but it's there now. :) Sep 08 22:38:13 well, you didn't change iTunes.... Sep 08 22:38:14 :) Sep 08 22:39:57 There's that I guess. :) Kinda my fault, though. I kept telling connecting machines I was an iTunes 4.0 server. Guess apple decided only to talk to iTunes 4.7 and above with the last release. Oops. Sep 08 22:41:05 Hi rpedde Sep 08 22:41:49 hey dyoung Sep 08 22:44:01 Can someone out there merge the heads? (I'm not in a position to be able to do this at present.) Sep 08 22:44:13 2 sec Sep 08 22:49:25 syncing Sep 08 22:50:04 well... Sep 08 22:50:20 except that I can't connect to monotone.nslu2-linux.org Sep 08 22:50:30 try monotone.vanille.de Sep 08 22:50:43 They are all equivalent... Sep 08 22:50:52 uh, yes. how do I switch server? Sep 08 22:51:13 Name it explicitly - monotone sync monotone.vanille.de org.openembedded.dev Sep 08 22:51:28 thx Sep 08 22:51:42 done Sep 08 22:52:58 looks good Sep 08 22:53:38 One thing that comes out of your GLIBC_2.0 problem - I found out how to fix my _call_via_r2 not found problem (libgcc was building it but not putting it into the version file). Sep 08 22:53:50 good :) Sep 08 22:54:09 My problem seems to have gone away anyways ;) Sep 08 22:55:37 GLIBC got scared with it heard that jb was on the case. Sep 08 22:56:55 good thing :) Sep 08 23:00:25 03byrontodd * 10unslung/make/cvs.mk: cvs.mk: Changed download to working OE named mirror. Sep 08 23:00:53 yay ;) Sep 08 23:04:03 ~hail NAiL Sep 08 23:04:06 * jbot bows down to NAiL and chants, "I'M NOT WORTHY!!" Sep 08 23:04:16 ~hail dyoung Sep 08 23:04:18 * jbot bows down to dyoung and chants, "I'M NOT WORTHY!!" Sep 08 23:05:48 03orepvik * 10unslung/Makefile: Sep 08 23:05:48 Makefile: Add a couple of packages to ds101 feed: Sep 08 23:05:48 ccxstream coreutils rtorrent cvs eggdrop elinks expat file flex Sep 08 23:05:48 gawk gdb gdbm gettext glib gnupg groff gzip joe logrotate lsof Sep 08 23:05:48 m4 make mc miau mt-daapd mtr mutt mysql nail nano nmap noip ntp openssh patch pcre Sep 08 23:05:49 proftpd psutils python rdate readline sane-backends sqlite sqlite2 strace syslog-ng Sep 08 23:05:51 textutils rtorrent unrar unzip usbutils wakelan Sep 08 23:06:10 * NAiL learns to count Sep 08 23:06:19 good lord, CIA-7 is needing some ex-lax after that Sep 08 23:06:48 * dyoung-web breaks out the castor oil Sep 08 23:08:39 rpedde, have you "de-underclocked" your nslu2 yet? Sep 08 23:10:48 ugh Sep 08 23:10:52 * NAiL goes to bed Sep 08 23:11:01 lol... not yet. Looks like something I need to do, though. Sep 08 23:12:37 it definitely helps mt-daapd... I rip everything VBR 192k and two streams could cause some "re-buffering" Sep 08 23:13:07 but with the clock mod, I've tested with 3 streams and zero re-buffering Sep 08 23:13:36 that's interesting... would probably help server-side transcoding, too. Sep 08 23:13:58 ogg decoding is pretty cpu intensive. Sep 08 23:14:22 jbowler: what was that about XScale fpu? Does the ixp have a fpu that we don't use? Sep 08 23:14:41 No, I got that wrong again - misreading gcc config files Sep 08 23:14:58 oh Sep 08 23:15:05 Got my hopes up there a sec Sep 08 23:15:26 Its an integer core only, for some reason some of the gcc configs seem to config in VFP support, but that's emulator support I believe. Sep 08 23:15:42 ah Sep 08 23:16:08 It will be interesting to see how a thumb mt-daapd performs... Sep 08 23:17:35 jbowler: thumb mt-daapd? Sep 08 23:17:43 indeed Sep 08 23:17:48 compiled for the thumb instructions, not the arm ones Sep 08 23:18:38 heh Sep 08 23:18:46 I see the load avg on the ds101 is... 1.00 Sep 08 23:19:42 more packages Nice. Sep 08 23:20:57 jbowler: advantage being what? smaller memory requirements? Sep 08 23:21:18 code size reduction at the cost of instruction count increase Sep 08 23:21:27 how big is the performance hit? Sep 08 23:21:39 Offset by clock-doubling the slug? Sep 08 23:22:21 impossible to say - if the instruction cache overflows then the performance benefit could be enormous, otherwise it depends Sep 08 23:23:09 oh! so you are saying that better cache localization could actually *improve* performance? Sep 08 23:23:11 libogg seems to come in at 86% of arm size, so that is 1.72 times slower in the worst case Sep 08 23:24:19 If the code being executed doesn't fit in the 16k (iirc) I-cache then there is a problem - libogg is 12.7kbytes and not all of that will be used Sep 08 23:25:28 thumb dhrystone with an arm library (string library in particular) goes from 3.0us per loop on ARM to 3.5us per loop with thumb Sep 08 23:25:59 That doesn't necessarily prove anything because gcc has specific hacks for dhrystone Sep 08 23:26:46 jbowler, really? ugh! Sep 08 23:27:06 pretty much every compiler does - they inline strcpy for small strings Sep 08 23:27:34 so dhrystone -Os -mthumb --> 3.0us, dhrystone -O3 -mthumb -> 2.2us Sep 08 23:27:55 fascinating Sep 08 23:28:32 and strcpy can be optimised for the specific string lengths dhrystone uses... Sep 08 23:30:05 oooh, mt-daapd is coming out at 77% of the arm size. Sep 08 23:30:24 significant Sep 08 23:30:31 178314 bytes -> 137510 bytes Sep 08 23:31:54 jacques: is jffs2 compressing files? Sep 08 23:37:05 jbowler, yes Sep 09 00:51:35 03koen 07org.openembedded.dev * r74c80fe1... 10/packages/avahi/avahi_0.4.bb: packages/avahi/avahi_0.4.bb: add 0.4, disable gtk and Qt Sep 09 01:40:21 ls -ltr tmp/deploy/ipk/ does not list any ipkg-native package. Where is this package? Sep 09 02:01:15 03koen 07org.openembedded.dev * rd657642f... 10/conf/distro/openzaurus-3.5.4.conf: conf/distro/openzaurus-3.5.4.conf: select glibc as provider for iconv and intl **** ENDING LOGGING AT Fri Sep 09 02:59:56 2005