| 00:08.59 | *** join/#edev menomc (n=amery@shell.opensde.net) |
| 00:11.25 | *** join/#edev landley (n=landley@rrcs-24-153-221-128.sw.biz.rr.com) |
| 00:11.25 | *** mode/#edev [+o landley] by ChanServ |
| 00:35.43 | landley | Morning all. |
| 00:36.14 | ds2 | morning |
| 00:37.06 | roxfan | night |
| 00:42.31 | *** join/#edev brolin (n=brolin@190.157.14.3) |
| 00:46.20 | solar | ds2: sorry. Yeah did that a good what seems 20times. |
| 01:15.26 | landley | Why am I banging on busybox again? |
| 01:15.44 | solar | cuz it's the orig swiss army knife |
| 01:16.12 | landley | cuz Vladimir emailed me about mdev (he wants to teach it module loading), and we started talking about the code, and that made me look at it. |
| 01:16.24 | landley | And now I _need_ to perform an #ifdefectomy just to make it STOP. |
| 01:16.26 | solar | thats fine. |
| 01:16.53 | solar | mdev is one of those apps that just needs to exist. |
| 01:16.57 | landley | I did enough work on toybox to know there's a better way to do most of the busybox infrastructure, but I don't want to try to push major design changes upstream into busybox, politically. |
| 01:17.03 | solar | I saw he started on SUBSYSTEM support |
| 01:17.09 | landley | Nor do I want to reinvent the wheel competing against a dozen or more active developers. |
| 01:17.27 | landley | Caught between a "don't wanna do that" and "don't wanna do that". |
| 01:17.29 | landley | :P |
| 01:17.53 | solar | dunno. But that applet is our little baby. |
| 01:17.59 | landley | Yup. |
| 01:18.16 | landley | And it's gotten tattoos and a few piercings and put on 40 pounds since last I saw it. |
| 01:18.24 | solar | recalls some guy posting the 1 mini-dev.c and I hacked it into mdev.c and ran with it and put in bb ;) |
| 01:18.39 | solar | err. You ran with it and put it in |
| 01:18.59 | landley | Frank Sorenson. |
| 01:19.11 | landley | I put a copyright notice for him at the top of the file, it's still there. |
| 01:19.18 | solar | good |
| 01:19.26 | landley | The mini-dev.c was just a naieve translation of the shell script I posted from bash to C. |
| 01:19.38 | solar | heh nice :) |
| 01:19.40 | landley | The shell script was from my 2004-era Firmware Linux. |
| 01:20.22 | solar | well some of the nuts and bolts that got added on semi ruined it. |
| 01:20.31 | landley | hmmm? |
| 01:20.45 | landley | Details? |
| 01:20.52 | solar | one mdev -s run will open/read/close /etc/mdev.conf for every device nod |
| 01:21.11 | landley | Ah yes. |
| 01:21.14 | landley | THAT'S TOTALLY BROKEN. |
| 01:21.16 | landley | Grrr. |
| 01:21.35 | landley | If I really start to care, I need to revert that. |
| 01:21.37 | ds2 | so it'd be a great benchmark for open/read/close syscalls? :D |
| 01:21.43 | landley | With a spoon and extreme prejudice. |
| 01:21.52 | landley | It was an mmap. There's a reason it was an mmap. |
| 01:22.10 | landley | They also do a lot of redundant strdup() of stuff that has acceptable lifetime rules to use directly. |
| 01:22.19 | landley | This code has clearly been maintained by people who don't understand it. |
| 01:22.33 | landley | My fault for leaving, in a way. Bruce #*%(&# Perens' fault for driving me away, though. |
| 01:23.01 | landley | They have #ifdef and curly bracket scopes crossing. |
| 01:23.09 | solar | well this is a good thing then. You can polish it up a bit and add better comments on "/* dont change this from mmap() cuz .... */ |
| 01:23.10 | landley | huddles in corner and whimpers for a bit. |
| 01:23.27 | landley | Looks like I need to reimplement the mmap stuff completely. Sigh. |
| 01:23.40 | landley | And the path creation/removal was done based on the 4k temporary buffer for a _reason_... |
| 01:23.51 | landley | Vladimir tells me he doesn't care about speed. |
| 01:23.56 | solar | I do |
| 01:24.20 | landley | I had _mainframe_ people using it because udev took 6 minutes to initialize their machine, and mdev took a little over 3 seconds. |
| 01:24.29 | landley | Oh well. |
| 01:25.04 | solar | gentoo used it on it's livecd cuz it was simple && fast |
| 01:25.20 | landley | It's apparently used in the initramfs of a normal gentoo boot, too. |
| 01:25.21 | solar | all linux livecd media that is. |
| 01:25.32 | solar | yep |
| 01:25.42 | landley | had to diagnose Mirell's server not booting earlier today, and the boot messages mentioned mdev. :) |
| 01:25.52 | landley | Turned out to be a problem with ipv6. |
| 01:26.01 | landley | is not a fan of ipv6. |
| 01:26.14 | landley | Dunno why it hung indefinitely waiting for ipv6 config after it had gotten ipv4 up and running. |
| 01:26.43 | landley | Apparently, if the local dhcp server doesn't hand out ipv6 information, gentoo will still be waiting for it the next day when you get around to figuring out why the server didn't boot. |
| 01:27.07 | landley | Do you have any idea why Denys is putting so much work into hush _and_ ash? |
| 01:27.25 | landley | Yay making busybox shells suck less, I'm just not sure why you'd make two of them bigger without trying to merge 'em... |
| 01:29.01 | solar | no idea. |
| 01:29.25 | solar | people seem to be pushing for ash posix compat cuz all the dash folks running around these days |
| 01:36.33 | landley | I have no objection to making ash posix compatable. |
| 01:37.01 | landley | If they added dir/{file1,file2} syntax and the >(command) <(command) redirects, I'd probably consider it a usable bash replacement. |
| 01:37.29 | landley | I realize that hush is at least theoretically nommu compatable. |
| 01:37.43 | landley | But the proper thing to do is make _ash_ work with nommu, and ditch hush. |
| 01:37.52 | landley | Also, get ash on a diet and make it so you can config the size down. |
| 01:37.59 | landley | Not have two major codepaths and make them both _bigger_. |
| 01:38.07 | landley | Oh well, not my project... |
| 01:39.06 | solar | I like it being two shells myself. |
| 01:39.29 | landley | Why? |
| 01:39.32 | solar | giz the nommu guys a playground which wont bloat the ash code with pretend vforks etc.. |
| 01:39.42 | landley | shrugs. |
| 01:40.03 | landley | I still think it should be possible to factor that sort of stuff out while leaving the majority of the code shared. |
| 01:40.31 | landley | It's extra work, but that's pretty much the busybox mission statement. |
| 01:41.38 | solar | honestly I sorta wish it was mostly a breakaway libash.so that handled all the syntax parsing and the applet in bb was pretty much just main() { while readline()... } |
| 01:42.02 | solar | not so much for busybox's needs. But having a nearly posix shell library parser would be handy |
| 01:43.05 | landley | solar: the while readline() approach turns out to be a _bit_ of an oversimplification. |
| 01:43.28 | landley | Your shell needs to be able to A) read extra lines for if statements, loops, here documents... |
| 01:43.45 | landley | And it needs to be able to give bits of lines _back_ and parse 'em later for ; && | and so on. |
| 01:43.54 | solar | yeah ;/ |
| 01:43.58 | landley | In fact you can combine 'em. |
| 01:44.20 | solar | I found that out when I attempted to give tclsh a proper readline() shell |
| 01:44.21 | landley | for i in `ls`; do echo $i; done |
| 01:44.38 | landley | I poked at bbsh/toysh for over a year. |
| 01:45.02 | landley | Got a pretty good idea of the problem space, but lost interest in actually _doing_ it. |
| 01:45.50 | landley | Same problem with my tinycc fork. Don't want to compete with the existing development community, but don't particularly want to work with 'em either. |
| 01:46.14 | landley | And thus code I'm interested in banging on rots. Oh well, plenty of other stuff to do... |
| 01:47.40 | solar | one thing we could really use is somebody looking over the security of busybox itself |
| 01:55.07 | landley | I used to do that. Unlikely to start again, though. |
| 01:56.43 | solar | we don't handle any netlink in mdev yet do we? |
| 01:57.03 | solar | http://www.securityfocus.com/bid/34536 <-- no risk that we could of made the same mistake right? |
| 01:57.30 | *** join/#edev kergoth (n=kergoth@ip24-255-7-190.tc.ph.cox.net) |
| 01:58.21 | solar | actually that lacks a lot of good info about the real problem |
| 01:59.03 | *** join/#edev g1powermac (n=g1powerm@unaffiliated/g1powermac) |
| 02:35.56 | *** join/#edev redblue (i=star@ppp078.108-253-207.mtl.mt.videotron.ca) |
| 02:54.17 | landley | solar: heh, I'd forgotten some of the original design elements of mdev. |
| 02:54.28 | landley | The reason it didn't do firmware loading is you call it from a wrapper script when you want that. Duh. |
| 02:54.41 | landley | I've been away from it for so long I've forgotten the basics... |
| 02:55.15 | landley | solar: we don't handle netlink. |
| 02:55.23 | landley | The thing about netlink is that it requires the sucker to run as a daemon. |
| 02:55.44 | landley | It's an optimization so that if 8 gazillion events come in via the hotplug path, you're not spawning 8 gazillion process to deal with it (and potentially having a thundering herd issue). |
| 02:56.08 | landley | But mdev -s doesn't require hotplug messages, so during boot it shouldn't matter so much. |
| 02:56.32 | landley | Keeping a daemon in memory is bad on embedded systems, we _want_ to have a process spawn and then _exit_ as quickly as possible. |
| 02:56.56 | landley | And embedded devices aren't mainframes, there aren't 1000 chatty devices attached to the system. We quiesce when not in use. |
| 02:57.23 | landley | solar: lacks _any_ info about the actual problem, from what I can tell. |
| 03:01.18 | keesj | also netlink works between the kernel and userland so I must be tricky to generate event without beeing root. (you can of course attach a usb thingy having weird code in the description) |
| 03:30.32 | solar | the addition of SUBSYSTEM will help as long as it's exported as an env var in mdev-event handlers |
| 04:27.51 | *** join/#edev mbuf (n=user@61.16.248.242) |
| 04:39.58 | *** join/#edev ZeZu (n=null@c-98-227-56-17.hsd1.in.comcast.net) |
| 04:56.11 | *** join/#edev dottedmag (n=dottedma@altlinux/developer/dottedmag) |
| 05:18.57 | *** join/#edev ZeZu_ (n=null@c-98-227-56-17.hsd1.in.comcast.net) |
| 05:58.54 | *** join/#edev onnoval (n=sam@123.252.146.52) |
| 08:10.15 | *** join/#edev mbuf (n=user@61.16.248.242) |
| 08:27.07 | *** join/#edev lyakh (n=lyakh@p57BD226C.dip0.t-ipconnect.de) |
| 08:38.14 | *** join/#edev roxfan2 (n=dunno@149.32-244-81.adsl-dyn.isp.belgacom.be) |
| 08:51.29 | *** join/#edev fwk (n=fwk@unaffiliated/fwk) |
| 09:04.11 | *** join/#edev cbrake_away (n=cbrake@oh-69-34-21-229.sta.embarqhsd.net) |
| 09:04.11 | *** join/#edev methril|fix_part (n=Methril@213.27.233.98) |
| 09:04.11 | *** join/#edev TheVoices (n=lou@c-68-53-134-58.hsd1.tn.comcast.net) |
| 09:04.11 | *** join/#edev file (n=file@asterisk/developer-and-muffin-lover/file) |
| 09:04.11 | *** join/#edev MonMotha (n=monmotha@72.86.127.225) [NETSPLIT VICTIM] |
| 09:04.11 | *** join/#edev wmat (n=btraynor@gromit.mixdown.ca) [NETSPLIT VICTIM] |
| 09:25.09 | *** join/#edev JoeLlama (n=snork@206.169.63.10) |
| 09:25.33 | *** join/#edev DrZimmerman (n=theo@77-56-240-146.dclient.hispeed.ch) |
| 10:09.25 | *** join/#edev petruss_der_6 (n=peterstu@vpn-148-241.vpn.uni-mannheim.de) |
| 10:09.56 | *** part/#edev petruss_der_6 (n=peterstu@vpn-148-241.vpn.uni-mannheim.de) |
| 11:43.13 | *** join/#edev florian (n=fuchs@f049104086.adsl.alicedsl.de) |
| 11:43.39 | *** join/#edev g1powermac (n=g1powerm@unaffiliated/g1powermac) |
| 12:00.00 | *** join/#edev devilsadvocate (n=devilsad@202.3.77.11) |
| 12:07.37 | *** join/#edev pirho (i=pirho@gateway/gpg-tor/key-0x2CEEC9CB) |
| 12:15.04 | *** join/#edev damunix (n=damunix@223.141.195-77.rev.gaoland.net) |
| 12:15.06 | *** join/#edev dijenerate (n=dijenera@64.210.41.166) |
| 12:29.26 | *** join/#edev likewise (n=chatzill@82-171-51-231.ip.telfort.nl) |
| 12:45.00 | *** join/#edev like2wise (n=likewise@82-171-51-231.ip.telfort.nl) |
| 12:57.29 | *** join/#edev GPSFan (n=kenm@64.92.145.112) |
| 12:57.29 | *** mode/#edev [+o GPSFan] by ChanServ |
| 13:06.44 | *** join/#edev onnoval (n=sam@123.252.146.52) |
| 13:07.32 | *** join/#edev onnoval (n=sam@123.252.146.52) |
| 13:08.18 | *** join/#edev onnoval (n=sam@123.252.146.52) |
| 13:09.46 | *** join/#edev g1powermac_PB (n=g1powerm@unaffiliated/g1powermac) |
| 13:20.01 | *** join/#edev brolin (n=brolin@190.157.14.3) |
| 14:37.39 | *** join/#edev devilsadvocate (n=devilsad@202.3.77.11) |
| 15:14.35 | *** join/#edev kergoth (n=kergoth@ip24-255-7-190.tc.ph.cox.net) |
| 15:32.44 | *** join/#edev pirho (i=pirho@gateway/gpg-tor/key-0x2CEEC9CB) |
| 15:33.44 | *** join/#edev like2wise (n=likewise@82-171-51-231.ip.telfort.nl) |
| 15:57.45 | *** join/#edev Othello (i=Othello@gateway/tor/x-aa33d83e3569df28) |
| 15:59.53 | *** join/#edev dijenerate (n=dijenera@64.210.41.166) |
| 16:11.46 | *** join/#edev Mirell (n=mirell@cpe-72-177-91-178.austin.res.rr.com) |
| 16:11.46 | *** mode/#edev [+o Mirell] by ChanServ |
| 16:47.17 | *** join/#edev Omegamoon (n=egoe@demolitionman.xs4all.nl) |
| 16:49.12 | *** join/#edev JoeLlama (n=snork@206.169.63.10) |
| 17:48.37 | devilsadvocate | anyone has any experience with TI TMS320 series DSPs and know if they are any good? |
| 18:18.23 | *** join/#edev lyakh (n=lyakh@p57BD226C.dip0.t-ipconnect.de) |
| 18:37.50 | *** join/#edev dijenerate (n=dijenera@64.210.41.166) |
| 18:37.52 | *** join/#edev flo_lap (n=fuchs@f049104086.adsl.alicedsl.de) |
| 20:27.00 | *** join/#edev Othello (i=Othello@gateway/tor/x-49c69d113707d15f) |
| 21:37.06 | *** join/#edev JoeLlama (n=snork@206.169.63.10) |
| 22:39.36 | *** part/#edev Omegamoon (n=egoe@demolitionman.xs4all.nl) |
| 23:04.39 | *** join/#edev like2wise (n=likewise@82-171-51-231.ip.telfort.nl) |
| 23:57.23 | *** join/#edev g1powermac_PB (n=g1powerm@unaffiliated/g1powermac) |
| 23:57.38 | *** join/#edev Dufflepod (i=Dufflepo@82-42-186-240.cable.ubr10.live.blueyonder.co.uk) |
| 23:57.46 | *** part/#edev Dufflepod (i=Dufflepo@82-42-186-240.cable.ubr10.live.blueyonder.co.uk) |