| 00:04.55 | CIA-8 | 03vapier * r12181 10uClibc/libc/string/sparc/: ignore *.os |
| 00:05.05 | CIA-8 | 03vapier * r12182 10uClibc/libc/stdio/Makefile.in: make cp silent |
| 00:15.25 | CIA-8 | 03vapier * r12183 10uClibc/libpthread/linuxthreads/sysdeps/sparc/ (5 files in 3 dirs): fix building by grabbing latest versions from glibc |
| 00:17.39 | *** join/#uclibc landley (n=landley@adsl-64-123-8-7.dsl.austtx.swbell.net) |
| 00:52.33 | CIA-8 | 03landley * r12184 10busybox/networking/ftpgetput.c: |
| 00:52.33 | CIA-8 | Patch from Tomasz Mateja: don't try to CWD to a _filename_, just send the full |
| 00:52.33 | CIA-8 | path to the STOR command. |
| 01:02.29 | landley | If I do an svn update, does that give me any new files that have been added to the repository? Or do I have to do something special? |
| 01:02.45 | landley | (I know it updates existing files, but when somebody does an svn add...) |
| 01:09.50 | *** join/#uclibc varun (n=varun@CPE-144-137-111-76.nsw.bigpond.net.au) |
| 01:12.10 | varun | building uClibc 2.4.20 with kernel 2.4.21 give error |
| 01:12.12 | varun | In file included from syscalls.c:1450: |
| 01:12.13 | varun | ../../../../include/linux/sysctl.h:391: error: conflicting types for 'NET_IPV6_ICMP_RATELIMIT' |
| 01:12.13 | varun | ../../../../include/linux/sysctl.h:386: error: previous definition of 'NET_IPV6_ICMP_RATELIMIT' was here |
| 01:12.13 | varun | make[4]: *** [_sysctl.o] Error 1 |
| 01:12.13 | varun | make[4]: Leaving directory `/home/varun/Desktop/myboot/uClibc-0.9.20/libc/sysdeps/linux/common' |
| 01:12.13 | varun | make[3]: *** [_dir_common] Error 2 |
| 01:12.15 | varun | make[3]: Leaving directory `/home/varun/Desktop/myboot/uClibc-0.9.20/libc/sysdeps/linux' |
| 01:12.17 | varun | make[2]: *** [_dir_linux] Error 2 |
| 01:12.19 | varun | make[2]: Leaving directory `/home/varun/Desktop/myboot/uClibc-0.9.20/libc/sysdeps' |
| 01:12.21 | varun | make[1]: *** [_dir_sysdeps] Error 2 |
| 01:12.23 | varun | make[1]: Leaving directory `/home/varun/Desktop/myboot/uClibc-0.9.20/libc' |
| 01:12.25 | varun | make: *** [_dir_libc] Error 2 |
| 01:12.27 | varun | any idea what to do? |
| 01:23.57 | landley | uClibc 2.4.20? |
| 01:24.25 | landley | Your kernel headers are horked. |
| 01:24.36 | landley | Lines 386 and 391 of the same file are conflicting. |
| 01:24.43 | landley | Not a uClibc problem, really... |
| 01:26.43 | varun | uclibc 0.9.20 |
| 01:26.50 | varun | kernel is 2.4.21 |
| 01:27.35 | landley | Is there a particular reason you're using a multi-year-old version that's 8 releases out of date? |
| 01:28.03 | varun | I tried with 2.6.14 even |
| 01:28.09 | varun | but it breaks |
| 01:28.28 | landley | With the 2.6 series you have to use Mazur's cleaned-up kernel headers, not the ones in the kernel itself. |
| 01:28.37 | varun | I need libata support for sata scsi in kernel |
| 01:29.03 | varun | which version you suggest |
| 01:29.08 | landley | http://ep09.pld-linux.org/~mmazur/linux-libc-headers/ |
| 01:29.23 | landley | If you want new features (like sata support), a new kernel's a good idea. |
| 01:29.35 | landley | What I'd recommend trying is the current uClibc against the current kernel. But that's just me. :) |
| 01:29.39 | varun | ohh |
| 01:29.58 | varun | but I dunno want to use buildroot |
| 01:30.05 | solar | uclibc-0.9.28 linux-2.6.11 and the above headers |
| 01:30.21 | solar | 2.6.12 and 2.6.13 both require nls support |
| 01:30.32 | solar | unless you do a bit of patching |
| 01:30.56 | varun | with 0.9.28 buildroot sucks |
| 01:30.57 | landley | I don't use buildroot. |
| 01:31.07 | landley | I made my own system, which is constructed via a big bash script. |
| 01:31.12 | landley | http://www.landley.net/code/firmware |
| 01:31.19 | landley | (This makes it easy to splice out the bits you need.) |
| 01:31.32 | landley | In fact, the first thing it does is generate a "/tools" directory (ala Linux From Scratch chapter 5.) |
| 01:31.36 | varun | thanks I'll take a look |
| 01:31.43 | landley | That's what all the sources/scripts/1* stuff does. |
| 01:32.00 | landley | Email me if you have any questions. rob@landley.net |
| 01:32.04 | landley | (Or if I'm still on here, of course... :) |
| 01:32.09 | varun | thanks for your help |
| 01:32.13 | landley | You're welcome. |
| 01:32.16 | varun | :-) |
| 02:00.35 | solar | landley: have you played with that miniudev recently? |
| 02:00.49 | landley | Not in the past few days, why? |
| 02:01.00 | landley | (The one posted to linux-kernel or the one I proposed writing?) |
| 02:01.06 | solar | yeah that one. |
| 02:01.12 | landley | The one posted to linux-kernel had no provision for specifying ownership or permissions. |
| 02:01.20 | landley | We need at least some way of specifying default values. |
| 02:01.37 | solar | somethign like this? ftp://tinderbox.x86.dev.gentoo.org/misc/mdev.conf |
| 02:01.39 | landley | We really _do_ need a config file to specify things, but should try to simplify it as much as possible and have sane defaults. |
| 02:02.50 | landley | Not bad. |
| 02:02.56 | landley | yeah, we could parse that. |
| 02:02.58 | landley | What's mdev? |
| 02:03.31 | solar | hacked up version of the miniudev you posted |
| 02:03.42 | landley | They move fast. :) |
| 02:03.51 | landley | Where's the source? |
| 02:04.15 | landley | I'm quite happy for somebody _else_ to write this. Is it integrated into busybox or standalone? |
| 02:04.21 | solar | same url ~/mdev.c |
| 02:04.26 | landley | (It's not like I haven't got enough on my plate already.) |
| 02:05.18 | landley | Very nice. |
| 02:05.33 | landley | shm and pts need to be in the config file. |
| 02:06.00 | solar | some way to say mkdir foo/bar/baz |
| 02:06.35 | landley | the buf[512] in get_device info needs to die. |
| 02:07.24 | landley | This is pretty cool. |
| 02:07.37 | landley | I need to spend more time on it when I'm not in the middle of 5 other things. |
| 02:08.07 | landley | What's cmp() for? |
| 02:09.36 | solar | I cant tell why it's really needed yet |
| 02:09.52 | solar | it's not like it's passed to any sorting |
| 02:09.57 | landley | Would you be ok integrating this into busybox? |
| 02:10.20 | landley | Or seeing this integrated into busybox and working based on the busybox version? |
| 02:11.58 | landley | (I'm tempted to clean it up tonight and check it in for 1.1.0-pre2.) |
| 02:12.29 | landley | Why is there a MAKEDEV symlink? |
| 02:12.37 | landley | (Isn't that script kind of obsolete in the udev world?) |
| 02:14.13 | landley | I'm tempted to add regexp support to the config file. Trying to figure out if it's a win or not... |
| 02:14.27 | landley | (It would avoid sr0 -> scd0 needing 16 repeats...) |
| 02:14.57 | solar | yeah but it was there for my testing |
| 02:14.59 | landley | It would have to be a config option. Statically linking in the regex engine is not fun. |
| 02:15.59 | landley | You have my email address? |
| 02:16.05 | solar | the overhead of regex adds nothing really |
| 02:17.08 | landley | Do you want me to bang on what you just sent me, or do you have more work pending for it? |
| 02:17.33 | landley | This looks highly promising and cool. |
| 02:17.41 | solar | well here is what I know. |
| 02:17.53 | solar | I dont like udev and devfsd is no more |
| 02:18.05 | landley | I've never managed to get udev set up. |
| 02:18.14 | landley | I've tried something like five different versions. |
| 02:18.26 | landley | Its configuration is arcane, the documentation sanskrit... |
| 02:18.37 | landley | And it keeps changing things like its basic invocation. |
| 02:18.52 | landley | You need two things out of it: an initial "populate /dev now" call, and a hotplug callback. |
| 02:19.03 | landley | And I have _no_idea_ how to get any release I've ever tried to even do the first one. |
| 02:19.41 | landley | I worked out a shell script from first principles, and Frank Sorensen's C program was a simple port of that shell script. |
| 02:20.09 | landley | I still have no idea what I'm doing, and the reason I haven't done more with this is I keep meaning to read through the current version of udev to see what the heck the requirements of this problem space are. |
| 02:20.20 | landley | That said, if you've got something that works for you, that's good enough for me to integrate into busybox. |
| 02:20.33 | landley | We can add more features as they're actually needed by real users. |
| 02:20.52 | solar | note it uses several static buffers of _POSIX_PATH_MAX |
| 02:21.08 | landley | I noticed. I'm trying to work out the best way to clean them up. |
| 02:21.10 | solar | it's 4K in size |
| 02:21.18 | landley | Those are implementation details. |
| 02:21.22 | landley | 4k in size is cool. :) |
| 02:21.51 | landley | Making the static buffers go away is sort of a security thing... |
| 02:22.06 | landley | (Not exactly, but close enough to make me uncomfortable.) |
| 02:22.15 | solar | is the use of asprintf() in bb applets ok? |
| 02:22.26 | landley | I'm trying to remember if libc or libbb has a "printf with malloc". |
| 02:22.31 | landley | And I think you just answered my question. :) |
| 02:22.40 | landley | If we use dprintf, we can use asprintf(). |
| 02:23.00 | landley | I think our rule of thumb is we use anything uClibc supports. |
| 02:23.25 | landley | That avoids the grottiest of the glibc internals, but not the documented extensions that actually make sense. |
| 02:24.59 | landley | If there isn't a libbb wrapper for asprintf() already, I'll make one. |
| 02:25.31 | landley | Its interface is kinda sucky. It should be char *malloc_printf(pattern, ...); |
| 02:25.43 | landley | returning NULL if it couldn't allocate, just like malloc... |
| 02:25.51 | landley | That's a side issue, of course... |
| 02:26.48 | landley | There is indeed bb_xasfprintf(const char *format, ...); |
| 02:26.51 | landley | Life is good. |
| 02:27.03 | solar | hack away :) |
| 02:27.14 | landley | You want me to take it from here? |
| 02:27.37 | landley | Tell you what: you bang on any outstanding things you still feel inclined to, then email me a copy of the *.c file and the *.conf file. |
| 02:27.40 | solar | sure go for it. I was play with it for kix |
| 02:27.49 | landley | I'll integrate 'em both into busybox. |
| 02:28.05 | landley | Or I can take the copies I just downloaded. Any pending requirements you think you'll need? |
| 02:28.21 | solar | next thing I was thinking was to make the '?' be a single char in an expression |
| 02:28.32 | solar | but if you use regex we are all good |
| 02:29.00 | landley | I think I'll add regex as an option. It does bloat a static link, and this is likely to be run from initramfs. :) |
| 02:29.10 | landley | config option, I mean. |
| 02:29.10 | solar | I did want to add the option to take username:group vs just uid/gid if %d:%d was %s:%s |
| 02:29.22 | landley | Lemme write this down... |
| 02:30.10 | solar | symlinks can be expressed as %s:%s perhaps vs the extra chars. |
| 02:30.25 | solar | tabs should work in place of spaces also |
| 02:30.37 | landley | I was wondering about how to support symlinks with spaces in them. |
| 02:31.01 | solar | dont do it :) |
| 02:31.16 | landley | Well that's an option too. :) |
| 02:32.21 | solar | not sure how to express dirs however. %s:%d |
| 02:32.30 | solar | like inittab maybe? |
| 02:32.31 | landley | I don't mind " -> " as the delimiter, since that's what ls outputs. |
| 02:32.58 | solar | yeah it was easy to script with awk a basic conf after the first time of running it |
| 02:33.50 | landley | strstr. |
| 02:34.14 | landley | " -> " is trivial with strstr, if I don't mind forcing exactly one space on each side (not tab). |
| 02:34.27 | landley | And if we go "hey, that's the format"... |
| 02:34.30 | landley | Document it. |
| 02:34.49 | solar | works for me |
| 02:35.19 | landley | Entry ending with / could be mkdir. Dunno if that's obvious enough, though... |
| 02:35.36 | landley | (It's easy to _parse_, I'm not sure it's easy enough to read.) |
| 02:35.46 | landley | If you assume people are going to read the documentation, you've already lost. |
| 02:36.01 | solar | people work from samples |
| 02:36.09 | landley | Exactly. |
| 02:36.21 | landley | Of course if the example file in busybox itself has a few comments, that's probably good enough... |
| 02:36.38 | landley | SLEEP=10 means what? |
| 02:36.58 | solar | I stuck the sucker in a loop and kept rm -rf /dev/ and changing the conf |
| 02:37.08 | solar | testing |
| 02:37.37 | landley | For the first version, numeric uid/gid is good enough. |
| 02:37.49 | landley | That can be upgraded in a later checkin... |
| 02:38.03 | landley | Is that really all we need? |
| 02:38.16 | landley | make symlink, mkdir, specify permissions? |
| 02:38.27 | solar | i think |
| 02:38.29 | landley | We need a shell trigger. |
| 02:38.39 | landley | The "hail mary pass" option. |
| 02:38.45 | solar | but if the user sets permissions via chmod we overwrite them |
| 02:38.46 | landley | If this thing shows up, shell out to this command line... |
| 02:38.58 | solar | which is probably really bad by having it that loop |
| 02:39.07 | landley | No, I think we make the device first (with our permissions), and then call the shell script? |
| 02:39.10 | landley | What loop? |
| 02:39.17 | solar | in main.c |
| 02:39.37 | landley | I think I'm yanking that. |
| 02:39.44 | solar | it should only really reset perms if huped |
| 02:39.47 | landley | mdev needs two modes. |
| 02:39.53 | landley | 1) populate an empty /dev, then exit. |
| 02:39.59 | solar | signal notify is what could send events |
| 02:39.59 | landley | 2) get called from /sbin/hotplug. |
| 02:40.23 | landley | I don't see it running as a daemon. |
| 02:40.35 | landley | It gets called to populate dev, and then the kernel calls hotplug whenever status changes. |
| 02:40.40 | solar | whats hotplugs role? |
| 02:40.45 | solar | oh |
| 02:40.57 | landley | Every time a device is added or removed from the system. (Or has a significant state change.) |
| 02:41.07 | solar | good thing you stopped me |
| 02:41.15 | solar | I was going to code it all in the same app |
| 02:41.16 | landley | You're still thinking in terms of devfsd. :) |
| 02:41.27 | landley | Oh it can still be in the same app. |
| 02:41.34 | solar | busybox |
| 02:41.41 | landley | Possibly "mdev -p" vs just "mdev". |
| 02:41.48 | landley | With -p being populate or something. |
| 02:41.53 | landley | mdev -s for startup, I dunno... |
| 02:42.06 | landley | Could even be two different names sharing most of the same functionality. |
| 02:42.46 | landley | The point is, udev isn't a daemon, and that design idea is good even if Greg KH's implementation isn't something that appeals to either of us. |
| 02:43.48 | solar | then bb is fitting |
| 02:44.01 | landley | ? |
| 02:44.05 | solar | kinda worried about bb however. it's overall size is starting to become big |
| 02:44.35 | landley | Yeah, it is. |
| 02:44.43 | solar | which starts to have it's own mem footprint for each execve of it |
| 02:44.49 | landley | But if you'll notice it's not that the individual apps are getting bigger. We're just getting more of them. |
| 02:45.01 | landley | They all CONFIG out. It just means "make allyesconfig" isn't your most intelligent option anymore. |
| 02:45.15 | landley | You may not need pidof or fuser or anything we added recently. |
| 02:45.23 | solar | yeah but I like having a backup of all the basic apps. |
| 02:45.35 | solar | they help in a pinch when you muck up your system |
| 02:45.46 | landley | If you're going to be greedy and want all the functionality we add, then you can't complain too much about the space that functionality takes up. :) |
| 02:46.08 | solar | nod.. I can/do config out what I dont want |
| 02:46.11 | landley | You can still build a pretty darn powerful busybox, statically linked against uclibc, in under 100k. |
| 02:46.19 | landley | Same here. |
| 02:46.28 | solar | anyway bb needs it now cuz there is really no other option for >=2.6.13 |
| 02:46.43 | landley | Yup. |
| 02:47.05 | landley | I'm trying to get busybox 1.1.0-final out by new year's. |
| 02:47.11 | landley | And this is very much a candidate for inclusion in that. |
| 02:48.11 | landley | Have you played with Matt Mackall's -tiny tree, by the way? |
| 02:48.27 | solar | I've poked at it |
| 02:48.43 | solar | it's a grapbag of features vs taking the entire thing in |
| 02:48.54 | solar | grab |
| 02:49.00 | landley | My Firmware Linux system's basic mode is a user mode linux kernel, plus uclibc, plus busybox (more or less makeallyesconfig). |
| 02:49.10 | landley | The end result is 2.5 megabytes. |
| 02:49.18 | landley | I'm trying to get that down. |
| 02:49.29 | landley | The kernel is over half of it... |
| 02:49.44 | landley | Busybox and uclibc are in a squashfs partition, you see. |
| 02:50.17 | solar | for what device? |
| 02:50.35 | landley | Well, the UML version runs as a standalone executable. |
| 02:50.47 | landley | User Mode Linux. |
| 02:51.18 | landley | It contains an initramfs that mounts the parent's root filesystem, looks for /proc/self/exe in that, loopback mounts its own executable at the appropriate offset to find the appended squashfs partition, and uses that as its root filesystem. |
| 02:51.34 | landley | Hence the entire OS is a single file. |
| 02:51.48 | solar | that could be handy |
| 02:51.55 | landley | I have a bootable version too (using a modified lilo that can load just the start of a kernel file if given a "length" parameter.) |
| 02:52.08 | landley | But I don't have an installer for the bootable version yet, so it's not useful to anybody else. :) |
| 02:52.16 | landley | And it's a serious pain to test. |
| 02:52.49 | landley | You can also append a gpg cryptographic signature to the end of the file, of course. |
| 02:53.00 | landley | And then your firmware updater can verify the signature... |
| 02:53.18 | solar | I like qemu cuz i can test my /usr/src/linux/arch/i386/boot/bzImage before I reboot |
| 02:53.25 | landley | And if it's ok it drops the file in the appropriate directory and calls LILO to add this firmware to the list. |
| 02:53.38 | landley | And then use LILO's -R option to provisionally boot the new firmware. |
| 02:53.54 | landley | Since -R sets the command line for the next boot only, if the firmware doesn't come up properly you just power cycle the device and get the old firmware back. |
| 02:54.02 | landley | If the firmware does come up, it sets itself as the new default. |
| 02:54.12 | landley | Much harder to turn the device into a brick that way. :) |
| 02:54.39 | landley | qemu is cool, but my ubuntu system doesn't have the SDL headers, and it's kinda hard to test stuff in am emulator that thinks it has no video card. |
| 02:54.46 | landley | Fixing that is a todo item... |
| 02:55.16 | solar | want a binary copy from another distro that works? |
| 02:55.34 | solar | or can live without till you fix it |
| 02:55.40 | landley | Tempting. |
| 02:55.54 | landley | But I have many to-do items and I'd rather fix this properly. |
| 02:59.17 | solar | are you going to bbify mdev right from the start? |
| 02:59.31 | solar | or make a standalone then bbify? |
| 03:00.47 | landley | bbify first. It's fairly easy. |
| 03:01.04 | landley | I'll probably thump on it later tonight, and then check it in. |
| 03:01.18 | landley | (That way, the version in svn will be up to date with each stopping point I come to.) |
| 03:01.28 | landley | Easier for other people to follow along and do their own development that way... |
| 03:01.38 | solar | local_errno and cmp can go also |
| 03:02.39 | landley | My "mdev.txt" file currently says regex support, username:group as string, zap cmp & local_errno, use bb_xasprintf() |
| 03:02.58 | landley | and add dir/ |
| 03:03.20 | landley | need to work out a syntax for shellout, though... |
| 03:04.39 | landley | Hmmm... Possible shellout syntax: |
| 03:04.57 | landley | devname uid:gid 777 $command line |
| 03:05.29 | landley | I.E. if after an entry for a device, there's a $ then the rest of the line is taken as a command line to run when that device changes. |
| 03:06.58 | landley | Sound vaguely usable? I don't plan to implement that part tonight, so the details can be worked out later. |
| 03:07.27 | landley | (Will there be any escapes in the command line to supply info, or do we just let them interpret the hotplug enviornment variables themselves?) |
| 03:07.46 | solar | copy the buf; step fwd 5 spaces and *p = 0; system(++p); perhaps |
| 03:07.52 | solar | s/5/3 |
| 03:08.02 | landley | That's the general idea. Make it trivial to code. |
| 03:08.18 | landley | We'd need to set an environment variable specifying which device name we just created, of course. (Well it could be a regex...) |
| 03:08.32 | landley | But I think it can stay pretty straightforward. |
| 03:08.53 | landley | Now I need to go look up what the various hotplug variables _are_... |
| 03:09.02 | landley | But first, let's get populating an empty /dev to work. |
| 03:09.07 | landley | First a few ground rules: |
| 03:09.24 | landley | 1) Is being hardwired to use /dev ok? |
| 03:09.26 | solar | another to the list. i think you need to be able to create user defined nods. |
| 03:09.32 | solar | like initctl is not in there |
| 03:09.33 | landley | (They can change the source code if they care.) |
| 03:09.56 | landley | initctl is a named pipe. |
| 03:10.29 | landley | In theory a shell script can create the named pipes, directories, and even symlinks. |
| 03:10.41 | landley | The advantage of having mdev.conf specify those is all the info about /dev is in one place. |
| 03:11.29 | landley | Having mdev create the actual /dev directory is silly. |
| 03:11.41 | landley | It should be a tmpfs mount, and I doubt mdev is calling mount for us. :) |
| 03:11.54 | landley | So I think mdev should assume that /dev exists and is empty when it starts. |
| 03:12.11 | landley | (Or if it's called again on an existing /dev, it just creates any missing device nodes and resets permissions on any that have changed.) |
| 03:12.45 | landley | I have the urge to call this thing uCdev, but that's a really ugly name. |
| 03:12.54 | landley | Stick with mdev. :) |
| 03:13.45 | solar | then it's only a few lines of code that need to change |
| 03:14.47 | landley | You know, ls -lp appends characters to filenames... |
| 03:15.07 | landley | Specifically, /=@| |
| 03:15.26 | landley | If we have directory support, and need named pipe support for initctl, then the logical thing to do is use those characters as indicators. |
| 03:15.40 | landley | Named pipe is |, directory is / |
| 03:16.16 | landley | In a way, " -> " is symlink. :) |
| 03:16.21 | landley | Dunno what = and @ are... |
| 03:17.49 | landley | I guess = is unix domain socket. |
| 03:17.55 | landley | (If that's what /dev/log is...) |
| 03:19.31 | landley | There are a lot of subdirectories. Do we care about these? |
| 03:19.37 | landley | (In /dev, I mean.) |
| 03:19.49 | solar | how about |
| 03:20.12 | landley | Things like /dev/snd |
| 03:20.13 | solar | <PROTECTED> |
| 03:20.29 | solar | the first char tells you what todo with the rest of the line |
| 03:21.32 | solar | that way you can parse easy using (*buf == action) && buf[1] == ':') |
| 03:21.36 | landley | Considering you can't have two devices with the same name, do we really need to pay too much difference to block vs char? |
| 03:22.04 | landley | strsep(" -> ") isn't too hard for finding symlinks. |
| 03:22.20 | solar | <PROTECTED> |
| 03:22.45 | landley | Hmmmm... |
| 03:23.21 | landley | That's not really any more obvious than the one that's there now. |
| 03:24.17 | solar | <-- trying to think out an easy way to take a /dev/ sanpshot conf and keep the bytes down |
| 03:24.27 | landley | What I'm worried about is checking that a line ends with / and getting tricked by a line "hda* 0:6 0660 $ls -l / |
| 03:24.46 | landley | Ok, one thing we've been forgetting: |
| 03:24.51 | landley | The ability to specify a _default_. |
| 03:25.13 | landley | Default uid, gid, and permissions for all devices. |
| 03:25.21 | landley | All devices otherwise unspecified, I mean. |
| 03:25.39 | landley | And actually, the mkdir and mkfifo commands also need uid:gid perm |
| 03:25.43 | *** join/#uclibc thraxisp (n=thraxisp@ottawa-hs-64-26-147-213.d-ip.magma.ca) |
| 03:25.54 | landley | So it's the _first_ entry on a line that we check if it ends with / or | |
| 03:25.57 | landley | That makes sense. |
| 03:26.43 | solar | yeah cuz you can strrchr and walk backwords on the string |
| 03:27.11 | solar | if thats where you want to set permissions and your worried about spaces |
| 03:27.26 | landley | Well, we already want to tokenize the string on whitespace. |
| 03:27.46 | landley | And if we're doing that then the signal for symlink is that uid:gid is "->" :) |
| 03:28.31 | landley | I'm looking at the subdirectories under /dev some more... |
| 03:29.05 | landley | Look at /dev/input |
| 03:29.13 | landley | It's /sys/class/input, only in its own subdirectory. |
| 03:29.17 | landley | Hmmm... |
| 03:31.28 | solar | rd,ida,pts,shm,cciss,ataraid is what I have off a livecd 2.6.x |
| 03:31.33 | solar | shm,shm/portage,shm/tinderbox,pts,misc,disk,disk/by-path,disk/by-id,disk/by-uuid,floppy,.udevdb,tts,ataraid,cciss,ida,input,rd another box |
| 03:31.57 | landley | shm is a directory that gets mounted, we don't maintain that. Same for pts. |
| 03:32.17 | landley | I notice that is 254:0, but /dev/hda1 is 3:1 |
| 03:32.31 | landley | Ahem. /dev/evms/hda1 is 254:40, but /dev/hda1 is 3:1 |
| 03:32.43 | landley | Once again, sirc's inability to cursor back and insert rears its ugly head... |
| 03:33.18 | landley | A lot of this is that people are using devfs naming schemes (ala /dev/loop/0)... |
| 03:33.59 | solar | loops were not created I noticed |
| 03:34.13 | solar | but the kernel supports 8 or so loopback devices |
| 03:34.43 | landley | Look at /sys/block/loop* |
| 03:34.52 | landley | Ours will create /dev/loop[0-7] |
| 03:35.00 | landley | Using the conventional naming scheme. |
| 03:35.15 | solar | hrmm |
| 03:35.19 | landley | That's the default behavior, anyway. whether the config should have some way of overriding that, we'll have to work out... |
| 03:37.07 | solar | looks like this kernel needs to be recompiled |
| 03:39.39 | solar | hrmm yeah let me hack on this a few mins if you have not started yet |
| 03:43.39 | solar | alot of the string handling is there cuz the program did not chdir() |
| 03:45.19 | landley | Ok, email me a copy when you want me to take over and I'll work off the version you email me. |
| 03:45.44 | landley | (Keeping in mind that I have other todo items for this evening and probalby won't get to it before tomorrow.) |
| 03:49.38 | landley | Off for food... |
| 04:12.07 | *** join/#uclibc landley (n=landley@adsl-64-123-8-7.dsl.austtx.swbell.net) |
| 04:19.21 | solar | landley: nevermind I gotta eat and hit the sack. the changes are trivial anyway |
| 04:58.24 | *** join/#uclibc landley (n=landley@adsl-64-123-8-7.dsl.austtx.swbell.net) |
| 05:41.54 | *** topic/#uclibc by landley -> edev |
| 05:42.02 | landley | Oops. |
| 05:42.22 | *** topic/#uclibc by landley -> uClibc - for discussion of uClibc and Busybox | uClibc |
| 05:42.22 | landley | <PROTECTED> |
| 05:42.22 | landley | <PROTECTED> |
| 05:43.02 | *** topic/#uclibc by landley -> uClibc - for discussion of uClibc and Busybox | uClibc 0.9.28 released 17 August 2005 | busybox 1.1.0-pre1 released 31 October 2005 | for non uClibc/busybox such as buildroot and general setup issues try #elinux or #edev |
| 05:43.22 | landley | Ok, how do I _view_ the topic on another channel? |
| 05:45.15 | solar | <PROTECTED> |
| 05:45.34 | landley | I figured it out. The # is important... |
| 05:45.43 | landley | How comes the mdev work? |
| 05:46.11 | solar | I broke it |
| 05:46.15 | solar | <solar> landley: nevermind I gotta eat and hit the sack. the changes are trivial anyway |
| 05:49.22 | landley | No biggie. :) |
| 06:11.53 | *** join/#uclibc landley (n=landley@adsl-64-123-8-7.dsl.austtx.swbell.net) [NETSPLIT VICTIM] |
| 06:11.53 | *** join/#uclibc thraxisp (n=thraxisp@ottawa-hs-64-26-147-213.d-ip.magma.ca) [NETSPLIT VICTIM] |
| 06:11.53 | *** join/#uclibc _cow (n=cow@M957P031.adsl.highway.telekom.at) [NETSPLIT VICTIM] |
| 06:11.53 | *** join/#uclibc mjn3-work (n=mjn3@h-66-134-116-110.dnvtco56.covad.net) [NETSPLIT VICTIM] |
| 06:11.54 | *** join/#uclibc mjn3 (n=mjn3@c-24-9-47-191.hsd1.co.comcast.net) [NETSPLIT VICTIM] |
| 06:11.54 | *** join/#uclibc zakalwe (i=zakalwe@you.sneakymonkey.net) [NETSPLIT VICTIM] |
| 06:11.54 | *** join/#uclibc rgb (i=fwuser@Quebec-HSE-ppp230809.qc.sympatico.ca) [NETSPLIT VICTIM] |
| 06:11.54 | *** join/#uclibc khem (n=khem@gateway-1237.mvista.com) [NETSPLIT VICTIM] |
| 06:11.54 | *** join/#uclibc CIA-8 (n=CIA@flapjack.navi.cx) [NETSPLIT VICTIM] |
| 06:11.54 | *** join/#uclibc minsuj (n=minsuj@211.232.190.6) [NETSPLIT VICTIM] |
| 06:11.54 | *** join/#uclibc Kaloz (n=kaloz@arrakis.dune.hu) [NETSPLIT VICTIM] |
| 06:24.33 | *** join/#uclibc wigyori (i=wigyori@fehercapa.kektintahal.pirospolip.hu) [NETSPLIT VICTIM] |
| 06:24.33 | *** join/#uclibc landley (n=landley@adsl-64-123-8-7.dsl.austtx.swbell.net) |
| 06:24.34 | *** join/#uclibc thraxisp (n=thraxisp@ottawa-hs-64-26-147-213.d-ip.magma.ca) [NETSPLIT VICTIM] |
| 06:24.34 | *** join/#uclibc _cow (n=cow@M957P031.adsl.highway.telekom.at) [NETSPLIT VICTIM] |
| 06:24.34 | *** join/#uclibc mjn3-work (n=mjn3@h-66-134-116-110.dnvtco56.covad.net) [NETSPLIT VICTIM] |
| 06:24.34 | *** join/#uclibc mjn3 (n=mjn3@c-24-9-47-191.hsd1.co.comcast.net) [NETSPLIT VICTIM] |
| 06:24.34 | *** join/#uclibc zakalwe (i=zakalwe@you.sneakymonkey.net) [NETSPLIT VICTIM] |
| 06:24.35 | *** join/#uclibc rgb (i=fwuser@Quebec-HSE-ppp230809.qc.sympatico.ca) [NETSPLIT VICTIM] |
| 06:24.35 | *** join/#uclibc khem (n=khem@gateway-1237.mvista.com) [NETSPLIT VICTIM] |
| 06:24.35 | *** join/#uclibc Kaloz (n=kaloz@arrakis.dune.hu) [NETSPLIT VICTIM] |
| 06:24.35 | *** join/#uclibc minsuj (n=minsuj@211.232.190.6) [NETSPLIT VICTIM] |
| 06:24.35 | *** join/#uclibc CIA-8 (n=CIA@flapjack.navi.cx) [NETSPLIT VICTIM] |
| 07:12.24 | *** join/#uclibc __cow (n=cow@M810P012.adsl.highway.telekom.at) |
| 07:16.21 | *** join/#uclibc cow (n=cow@M872P010.adsl.highway.telekom.at) |
| 07:24.06 | *** join/#uclibc landley (n=landley@adsl-64-123-8-7.dsl.austtx.swbell.net) |
| 07:25.22 | CIA-8 | 03psm * r12185 10uClibc/Makefile.in: Remove copied sources |
| 08:42.52 | *** join/#uclibc vrm (n=vrm@248.55.101-84.rev.gaoland.net) |
| 08:53.19 | *** join/#uclibc psaksa (n=pate@ip212-226-134-137.adsl.kpnqwest.fi) |
| 10:21.15 | *** join/#uclibc CIA-10 (n=CIA@flapjack.navi.cx) |
| 10:45.26 | *** join/#uclibc CIA-2 (n=CIA@flapjack.navi.cx) |
| 11:50.46 | *** join/#uclibc prpplague (n=billybob@200.50.80.24) |
| 13:02.35 | *** join/#uclibc sjhill (n=sjhill@eth13.com-link.com) |
| 13:22.05 | *** join/#uclibc CIA-2 (n=CIA@flapjack.navi.cx) [NETSPLIT VICTIM] |
| 13:22.05 | *** join/#uclibc psaksa (n=pate@ip212-226-134-137.adsl.kpnqwest.fi) [NETSPLIT VICTIM] |
| 13:22.05 | *** join/#uclibc wigyori (i=wigyori@fehercapa.kektintahal.pirospolip.hu) [NETSPLIT VICTIM] |
| 13:22.06 | *** join/#uclibc thraxisp (n=thraxisp@ottawa-hs-64-26-147-213.d-ip.magma.ca) [NETSPLIT VICTIM] |
| 13:22.06 | *** join/#uclibc mjn3-work (n=mjn3@h-66-134-116-110.dnvtco56.covad.net) [NETSPLIT VICTIM] |
| 13:22.06 | *** join/#uclibc mjn3 (n=mjn3@c-24-9-47-191.hsd1.co.comcast.net) [NETSPLIT VICTIM] |
| 13:22.06 | *** join/#uclibc zakalwe (i=zakalwe@you.sneakymonkey.net) [NETSPLIT VICTIM] |
| 13:22.06 | *** join/#uclibc rgb (i=fwuser@Quebec-HSE-ppp230809.qc.sympatico.ca) [NETSPLIT VICTIM] |
| 13:22.06 | *** join/#uclibc khem (n=khem@gateway-1237.mvista.com) [NETSPLIT VICTIM] |
| 13:22.06 | *** join/#uclibc minsuj (n=minsuj@211.232.190.6) [NETSPLIT VICTIM] |
| 13:22.06 | *** join/#uclibc Kaloz (n=kaloz@arrakis.dune.hu) [NETSPLIT VICTIM] |
| 13:53.15 | *** join/#uclibc psm (n=mps@host-6.mikroweb.hu) |
| 14:31.52 | *** join/#uclibc vapier (i=UserBah@wh0rd.org) |
| 14:32.34 | prpplague | morning folks |
| 15:15.42 | *** join/#uclibc _cow (n=cow@M975P001.adsl.highway.telekom.at) |
| 15:38.04 | *** join/#uclibc thraxisp (n=thraxisp@ottgate.precidia.com) |
| 16:46.33 | psm | mjn3: is #define _STDIO_UTILITY needed ? I can see only #define of it, but no use |
| 17:00.20 | CIA-2 | 03psm * r12186 10uClibc/libc/ (4 files in 3 dirs): Remove #define _STDIO_UTILITY, can't find any use of it |
| 18:00.58 | *** join/#uclibc JockeHome (n=JockeHom@84-217-87-110.tn.glocalnet.net) |
| 18:02.36 | psm | JockeHome: hi |
| 18:06.38 | JockeHome | hi |
| 18:06.57 | psm | I think glibc uses weak_alias because they have many versions around that overwrite each other (generic/os-specific/arch-specific/os-arch-specific) |
| 18:07.29 | psm | we do not have that problem because we build only one version based on the chosen options |
| 18:07.33 | JockeHome | I see, maybe uClibc should do the same? |
| 18:07.47 | JockeHome | Would that solve IMA for you? |
| 18:07.53 | psm | generic->generic-opt->arch-opt |
| 18:08.15 | JockeHome | ? |
| 18:08.19 | psm | IMA is not related to this, for IMA I have to split most of the MSRC files |
| 18:08.45 | psm | uClibc (at least libc/string) has those options |
| 18:09.09 | psm | generic->opt->arch-opt |
| 18:10.28 | JockeHome | OK, I am that familar enough with the build system for string |
| 18:10.50 | JockeHome | s/that/ not that/ |
| 18:11.37 | psm | I have saved some compile time by excluding all the duplicate objects before compile |
| 18:11.49 | psm | bon appetite |
| 18:12.02 | JockeHome | :) |
| 18:13.30 | JockeHome | need to eat some more, back later |
| 18:26.29 | CIA-2 | 03psm * r12187 10uClibc/include/signal.h: Correct a problem introduced by porting the glibc header |
| 18:27.06 | *** part/#uclibc nitinkg (n=ngupta@gateway-1237.mvista.com) |
| 18:41.50 | *** join/#uclibc nitinkg (n=ngupta@gateway-1237.mvista.com) |
| 19:42.38 | *** join/#uclibc andersee (n=andersee@codepoet.org) |
| 20:02.42 | vapier | mjn3-work: alive ? i'd like to punt the mips/bits/types.h ... the only thing of concern though is __kernel_ipc_pid_t |
| 20:03.30 | vapier | mips/bits/types.h says 'typedef long int __ipc_pid_t' but mips/bits/kernel_types.h says 'typedef int __kernel_ipc_pid_t' |
| 20:03.44 | vapier | should the kernel_types be changed to 'long int' or can i just punt types.h and forget about it ? |
| 20:36.37 | *** join/#uclibc kergoth (n=kergoth@c-24-118-219-25.hsd1.mn.comcast.net) |
| 23:00.10 | *** join/#uclibc khem (n=khem@gateway-1237.mvista.com) |
| 23:14.53 | psm | vapier: the kernel (2.4) has itself int, unsigned short and short depending on arch |
| 23:15.38 | psm | 2.6 seems missing this |
| 23:16.31 | *** join/#uclibc cow (n=cow@M942P004.adsl.highway.telekom.at) |
| 23:17.45 | vapier | ? |
| 23:21.03 | psm | grep kernel_ipc_pid_t include/asm-*/types.h |
| 23:24.23 | vapier | i remember there was a lot of ipc crap we had to tinker with before 0.9.28 which is why i'm asking mjn3 about it for mips :) |