irclog2html for blob on 2002.01.07

10:32:52erikm'lo
10:43:53Russmorning
10:44:06Russmd5, or rc5?
10:44:12erikmoops
10:44:13erikm:)
10:44:22erikmlol :)
10:44:25Russwoohoo, encyrption
10:44:39erikmnah, just an md5 hash
10:44:42RussI'm sure someone's boss is asking for it
10:44:51Russ"is our bootloader encrypted?"
10:45:01erikmthat's why I made it configurable :
10:45:21erikmbtw, I'm thinking about doing a 2.0.5-pre2 snapshot release
10:46:09erikmthe code is relatively stable right now, and I think we're going to see lots of changes
10:47:08Russbbiaw
10:47:10erikm(I'm working with BZFlag and somebody else from lineo to port it to non-SA architectures as well)
12:56:20seletzhi all!
12:56:38seletzHappy new Year to you! (a bit late, tough ... :)
13:13:43erikmre
13:17:33erikmseletz: happy newyear
13:17:55erikmseletz: btw, I'm thinking about doing a new development snapshot today
13:18:10seletzerikm: hi!
13:18:54erikmseletz: any good or bad feelings about that?
13:18:57seletzerikm: well, I'm currently solving some cvs conflicts i got today.
13:19:09erikmheh :)
13:19:34seletzerikm: but they seem mostly due to lack of recent check-outs :)
13:19:41erikmbtw, system3 doesn't compile in the CVS version.
13:19:44erikmcvs update -dAP
13:19:55seletzerikm: :/
13:20:17erikm(I have "update -dP" in my .cvsrc)
13:20:18seletzerikm: what's about pcmcia? I noticed some changes in diag/system3.c
13:20:32seletz(me too.)
13:20:54erikmseletz: that was because Chris Hoover put the SA1111_BASE in the arch dependent files
13:21:17erikmseletz: the compiler complains about missing declaration of function __REGP
13:21:25seletzerikm: thats a good idea anyway.
13:21:25erikm(line 117 in sa1111_init)
13:21:39seletzchecking out
13:22:00seletzyes, i know.
13:22:01erikmand it complains about SA1111_VBASE being undeclared
13:22:06seletzyes, i know.
13:22:14erikmgot the feeling that those two errors are related
13:22:34seletzthats because Chris did not copy all #defines to system3.h
13:23:04seletzerikm: thats all because the sa1111.h in linux kernel code changed
13:23:04erikmoh, and I also updated some of the serial stuff in system3.c
13:23:10erikmah, right
13:23:39seletzerikm: i make it working (i.e. compilable at last) and do a checkin.
13:24:03seletzerikm: could you wait with the dev snapshot?
13:24:19erikmsure
13:24:26seletz(while i resolve the errors)
13:24:29erikmbtw, xmodem is *fast*
13:24:30seletzok
13:24:40seletzah, the topic.
13:25:06erikmespecially if you tell sx to use 1kB blocks with the "-k" option: 10080 bytes/s
13:25:07seletzwell, xmodem up/download is really nice.
13:25:24seletzwell, thats fast!
13:25:37seletzwhat about md5?
13:25:50erikmit normally uses 128 bytes blocks, and then it's about 6200 bytes/s
13:25:52seletzcan we md5 check arbitary memory regions?
13:25:59erikmyes, you can
13:26:02seletz:))
13:26:37seletzsolving conflicts
13:27:20erikmhmm. I have some minor patches in my tree, I'll commit them as well
13:28:27seletzerikm: well, I'll call this day a cleanup and patching day. I'll remove some cruft from my code as well.
13:36:17seletzerikm: well, sa1111.h is my fault. i never did a checkin with SA1111_VBASE defined. doh.
13:45:28seletzerikm: have you seen the "regs" command in system3.c?
13:46:06seletzerikm: maybe we should move it to a non arch dep source file?
13:47:03erikmlooks
13:47:09Sammylater to join :)
13:47:20Sammygo home ....
13:48:46erikmseletz: looks like a good plan
13:49:25erikmseletz: BTW, the names of those registers is in SA1100.h
13:49:49erikmseletz: (and be careful that the SA1100 has slightly different registers)
13:50:03seletzerikm: ok, i'll take a look and expand the list. Where to put it in, then?
13:50:17erikmsrc/diag/regs.c
13:50:23seletz:)
13:55:04seletzerikm: I'll create a include/blob/regs.h with some #defines foe SA1100 and SA1110. ok?
13:55:54erikmseletz: like what?
13:56:21erikmseletz: everything is already defined in include/asm-arm/mach-sa1100/SA-1100.h in the kernel source
13:56:41seletzerikm: like #if defined(SA1110)define the list of reg names for StrongARM #endif and so on
13:57:04seletzerikm: because of diffs between sa1100 and sa1110 etc.
13:57:17erikmseletz: better put that in regs.c, no need to export that to other files
13:57:35erikmseletz: btw, make it regs-sa11x0.c
13:57:45seletzerikm: ok. less work, too :)
13:57:46seletzok
13:58:38erikmseletz: there is a non-StrongARM port in the work, so I'd rather have such architecture dependent stuff in files that can be clearly distinguished
13:58:55erikmseletz: (that's btw exactly the reason why I cleaned up the serial driver)
14:03:13seletzerikm: cotulla?
14:03:41erikmseletz: no, ARM 720T based. and there's also somebody working on ARM7TDMI
14:04:15seletzerikm: man, arm gets crowded, does it?
14:04:42erikmseletz: it already was. arm is everywhere
14:04:56erikmseletz: your harddrive probable has an ARM core
14:06:07seletzerikm: hmm, running linux on my hd holding linux and built in in a linux box? Oh, wait, hmmm
14:06:17seletz:)
14:07:23erikmseletz: there are people thinking about running armlinux on their GigE cards :)
14:12:31seletzerikm: ok, can we put printf() from brad's code int libblob?
14:13:01seletzerikm: this would allow a lot of cleanup in blob
14:13:10erikmseletz: I intend to do so, but I'd rather first release 2.0.5-pre2
14:16:00seletzerikm: ok, in that case i'll have to wait with my patches too, because they are pcmcia related, and they use printf(). So i'll hack up a regs-sa11x0.c and submit it, then change system3.h to define SA1111_VBASE and that's it. The rest after the pre2 release. agreed?
14:16:31erikmseletz: agreed
14:17:43erikmreboots to test a VM patch for riel
14:43:28erikmreturns
14:50:24seletzerikm: hmm, we currently don't have a #define which holds the cpu type info. I'll add a #define in config.h and configure.in. Ok?
14:50:50erikmseletz: I already had a patch doing that :)
14:51:12erikmwait a minute....
14:51:16seletzerikm: :)
14:55:42erikmcvs commits
14:57:10erikmif you wait a for a second I'll also commit a minor patch for src/blob/{Makefile.am|rest-ld-script.in}
14:57:23seletznp
14:59:08erikmok, done
15:01:15erikmyou might need to run "make clean" if src/blob/blob doesn't link
15:03:40seletztrying out
15:06:32seletzerikm: hmmm, "CPU" is #undef'd in config.h. Should'nt CPU set in configure.in?
15:06:48erikmrun tools/rebuild-gcc
15:06:54seletzdoh
15:07:33erikmand use the --enable-maintainer-mode flag for configure
15:11:55seletzerikm: well, i _did_ use maintainer mode. and in my configure.in there's no AC_DEFINE/SUBST whatever filling CPU. (i did a make clean and configure too.)
15:13:19seletzprpplague: hi & happy new year ...
15:15:13erikmhell prpplague
15:15:27erikms/hell/hello/ <-- freudian mistake :)
15:15:44erikmit is in configure.in
15:15:53erikmseletz: try cvs update -dAP configure.in
15:16:31erikmseletz: you might have a sticky tag on configure.in
15:16:57prpplagueerikm: morning
15:20:24erikmremembers there are free drinks downstairs
15:20:45erikmbbl
15:21:32prpplagueerikm: later dude
16:57:16erikmreturns
16:58:16erikm*burps*
16:58:25erikmI don't think I need something to eat RSN
16:58:26erikm;)
17:24:50erikmseletz: src/diag/system3.c still doesn't compile for me. which kernel are you using?
17:31:03akirahi
17:31:08seletzhi too
17:31:11seletz:)
17:31:40erikmakira: I removed include <blob/sa1111.h> from regs-sa11x0.c, it breaks all other architectures and isn't needed
17:31:59erikms/akira/seletz/
17:32:06akiraPONG
17:32:15akiraor whatever :)
17:32:42seletzerikm: oops. yes, of course.
17:33:31seletzerikm: made it not configurable, i'm too lazy now.
17:33:50erikmdoesn't matter
17:34:37seletzhmm, i'm using a 2.4.16-rmk2
17:34:48erikmchecks
17:35:17erikm2.4.13-ac5-rmk2. that's probably the problem
17:35:17seletzchecking too
17:35:40erikmI needed a known stable kernel for a couple of demos
17:35:41seletzerikm: yes. they changed somewhere in between. sigh.
17:36:23seletzerikm: can you quickly switch over?
17:36:28erikmseletz: not really
17:36:34seletzerikm: hmmm.
17:36:54erikmseletz: could you run a "make dist", untar the tarball in /tmp, configure for system3 and see if that compiles?
17:37:21seletzerikm: np
17:37:37seletzworking
17:38:03seletzerikm: i'll take a fresh cvs version.
17:39:10erikmseletz: the "make dist" tarball is interesting, it makes it clear if you forgot to add a file to a Makefile.am
17:39:30seletzwe'll see ...
17:44:23seletzbuilds, except for diag, which doesnt build because of system3.c. Argh.
17:44:32seletzchecking out why
17:45:43erikmseletz: it might be an include file with a sticky CVS tag
17:46:35seletzone moment
17:47:12seletzdoh!
17:47:14seletz:)
17:47:30erikmsomething trivial? ;)
17:47:46seletzgood test, make dist....
17:47:53erikm*nod* :)
17:48:07akira;)
17:48:30BZFlagerikm: what's the status on sa1111 support?
17:48:42erikmBZFlag: seletz knows
17:48:50BZFlagok, seletz? ;-)
17:48:55akirai'll expect u to teach me all this seletz :)
17:49:08erikmBZFlag: I'm going to do a 2.0.5-pre2 development snapshot tonight
17:49:22seletzworking
17:49:24seletz:)
17:49:39BZFlagwhich platform uses it now? and for what?
17:49:55seletzerikm: we'll see if this commit breaks it further more or solves it ...
17:50:02seletzon moment pls
17:50:13erikmrolls drums
17:50:17BZFlaglooks like I need to add it to get at the sa1111 gpio pin to enable flash writes on this silly piece of hardware.
17:50:40erikmBZFlag: the idr?
17:51:01BZFlagyep.
17:51:38BZFlagone intel 16MB flash chip with write enable tied to a sa1111 gpio. nice, huh?
17:51:54erikmthat's another way to do it, yes :)
17:52:01seletzwas a sticky tag in sa1111.h
17:52:14erikmseletz: ah, that explains
17:52:18BZFlagand 2 other gpio on the sa1111 do lcd and backlight power.
17:52:56BZFlagcourse I think there are more than 3 free gpio's on the 1110.
17:52:57erikmBZFlag: well, that's possible as well. ipaq is not clean either
17:52:58seletzerikm: ok, i'll test it all again
17:52:59BZFlagargh.... ;-)
17:53:23erikmBZFlag: especially their egpio register that's *write* *only* :(
17:54:04BZFlagheh
17:54:08seletzerikm: doh, those write only horrors i got thankfully out of the head of my hw design guy ...
17:54:39erikmseletz: good. write only registers are evil
17:54:46BZFlagagreed!
17:54:51seletzyep
17:55:21BZFlagspecially when they map the bit someplace else for reads. ugh.
17:55:47erikmBZFlag: yeah, the VGA did that. that was really hell
17:55:58seletz(they wanted to mimic a serial io -> shift reg to save some lines. they wanted to make this reg write-only. write-and-forget. ARGH! i said.)
17:55:58erikmBZFlag: and also hell to emulate
17:56:31erikmBZFlag: (/me wrote the VGA emulator for DOSemu)
17:58:05erikmwrites release notes
17:58:09seletzerikm: wow, it builds.
17:58:12seletz:)
17:58:18seletz_phew_
17:58:19erikmgood :)
17:58:27seletzsorry, tough.
17:58:47seletzBZFlag: ok, now sa111 questions?
17:58:53seletzs/sa11/sa111/
17:58:57seletz:)
17:59:07seletzf**ed it up
17:59:52seletzBZFlag: sa1111 is used on system3 for LCD and PCMCIA
18:01:23BZFlagok,
18:04:40seletzBZFlag: the only thing is that ATM its only in diag.
18:04:48erikmseletz: btw, I think you need this patch in smctest() in src/diag/system3.c:
18:04:53erikm-u32*base= (u32 *)SMC_BASE;
18:04:54erikm+u32base= (u32)SMC_BASE;
18:05:16seletzchecking
18:06:27seletzwell, quite redundant IMHO.
18:06:38erikmyeah, but removes two compile time warnings
18:06:44seletzok
18:06:56seletzapplied
18:06:58seletz:)
18:12:16erikmseletz: did you commit it to CVS?
18:12:46seletzoh, sorry. one moment
18:13:48seletzdone.
18:14:01erikmnotes that the MD5 digest for a file uploaded with xmodem differs from the file on the host
18:14:07erikmseletz: thx
18:14:20seletzoops. that's not good ...
18:14:48seletzoff-by-one error in checking area?
18:14:55erikmah, I see why. xmodem uses 128 byte packets and stuffs data at the end of the file
18:15:02erikmtries
18:15:35erikmok. to make it good, first do dd if=blob of=blob-for-md5 bs=128 conv=sync
18:15:59seletzdd is _super_
18:16:10erikmnow md5sum blob-for-md5, and you'll see that the md5sum is correct
18:16:20erikmwe need ymodem or zmodem :)
18:17:02seletzwell, we need tftp i'd say
18:17:06seletz:))
18:17:10erikmIIRC ymodem is almost the same as xmodem+1k blocks+extended CRC and support for filenames and file length
18:19:08erikmfrom the release note: "We admit, we made a mess of it." :)
18:22:04BZFlagseletz: well, I'll have to move/copy it over to main for this ugly board hack.
18:22:29erikmBZFlag: if you do so, put it in src/lib/
18:22:52erikmBZFlag: (but please wait for a couple of hours until I release 2.0.5-pre2)
18:23:38BZFlagagreed
18:23:43erikmBZFlag: btw, any news from duck?
18:24:15BZFlagyou got his emails last week, yes? nothing since. (I'm still at home)
18:24:38erikmoh, wait, you're at pacific time
18:24:49BZFlagit might be nice to make it a config option to move stuff back and forth tween diag and blob proper.
18:24:53BZFlagmountian.
18:25:28erikmBZFlag: I'd rather keep blob as small as possible
18:26:10erikmBZFlag: if we need to sa1111 support in order to get some platforms supported in blob, I'd rather move the support in src/lib/
18:26:27seletzerikm: but if BZFlag needs sa1111 support, we'll have to move things over
18:26:42seletzerikm: just a copy hack is not suitable IMHO
18:26:49erikmseletz: yes, and because you need it in diag, we'd rather put it in src/lib/
18:26:55seletzyup
18:27:01erikmseletz: just to avoid code duplication
18:27:08seletzyup**2
18:27:13erikm:)
18:28:14seletzerikm: just to give you a shock: if someone ever wants someting like a custom splash screen in blob, we'll have to move lcd.c as well.
18:28:18erikmOK, so on what platforms does this blob version work? Assabet, LART, System3, IDR?
18:28:41erikmseletz: well, I'm thinking about some kind of splash screen support. configure time option of course.
18:28:50seletz:)
18:28:50BZFlaghence me "would be nice" statement.
18:29:10seletzgenau, erm, yup
18:29:29BZFlagif all platform code that works, but is not explicitly selected for "blob" got put in "diag" that would suite me just fine.
18:29:57erikmBZFlag: doesn't always make sense
18:30:16seletzerikm: if libblob gets big, thats not a problem, is it? blob links against it, and if the linker doesnt need some code it doesnt get in, right?
18:30:30erikmseletz: right
18:30:40BZFlagagreed.
18:31:06BZFlagso perhaps it's all in "diag" and only selected stuff in "blob"?
18:31:11seletzerikm: so, if we just make _one_ place to add commands and support funcs, then it would be libblob
18:31:34BZFlagseletz: you still need to tell "blob" which things to link in.
18:31:38erikmseletz: the linker works on a per file base for libraries. that's why I try to put functions in separate files (strncpy(), for example)
18:31:39seletzerikm: ATM its all a mess when i for ex want peek and poke in blob _and_ diag
18:31:41erikmBZFlag: *nod*
18:31:51erikmBZFlag: that's done in initcalls.c, for example
18:32:34seletzerikm: so make every command or commandset in sepparate files and put them in libblob
18:32:44seletzs/sepparate/separate/
18:32:50erikmseletz: move the peek() and poke() implementation to src/lib/peek.c and src/lib/poke.c
18:33:01seletzthat's what i mean
18:33:29erikmseletz: now make a selectively compiled src/blob/peek.c that only defines the __command() and it automatically gets linked in
18:33:32seletzerikm: same for upcoming pcmcia, ide and filesystem stuff
18:33:37erikmright
18:34:27seletzok, then we could make a blob/commands.c where the __command() calls live. no need for separate files under blob and diag
18:35:00erikmthere already is blob/commands.c :)
18:35:13seletzyou know what i mean :)
18:35:35seletzname it blob/this_is_where_the_command_macros_are.c
18:35:40seletz:D
18:35:55erikmyes, but they would become a #define hell
18:36:11erikm#ifdef BLOB_CONFIG_WANT_COMMAND_FOOBAR
18:36:18seletzyup, agreed. but better one hell than more than one ...
18:36:21erikm__commandlist(foobar, "foobar", foobarhelp);
18:36:23erikm#endif
18:37:00seletzerikm: one could even come up with a shell/perl/bash script which generates this file
18:37:28erikmgenerated files... cool
18:38:39seletzerikm: no more hell but one single place to add/remove commands.
18:38:51erikmthinks about it
18:39:10seletzerikm: _and_ this file could be generated into src/diag and src/blob with the same script and syntax
18:40:50seletzerikm: the script could even automagically grep through the src/lib/*.c source for special tags to get help strings coherent, and check for commands etc
18:41:09seletzgetting too excited
18:45:00akira:)
18:59:17seletzerikm: what did you come up with?
18:59:21seletz:)
19:00:08erikmstill writes release note
19:00:09erikms
19:00:20erikmand also slightly updates the README file
19:00:35seletz:)
19:01:03erikmDo you think I should mention the #blob IRC channel in the README file?
19:01:16erikmthere is a risk that we get swamped with people over here
19:01:35seletzyes, i did mention it in some emails already.
19:01:48erikmthat's not really a problem
19:01:59seletzif we get crowded we could open a #blob-devel or something
19:02:09seletz(private for developres)
19:02:32erikmbut I'm a bit afraid that we'll reduce this channel to a blob helpdesk, and that's not really my intention
19:02:48seletzhmmm
19:03:22seletzthats a decision which should taken with care, IMHO.
19:03:38erikmI'll just don't mention it right now
19:03:49akirai'm quiet don't kick me :)
19:04:02akirabesides i'm really nice in trhe morning :)
19:04:05seletzwe could give blob a boost in popularity.
19:04:18seletzand thus get more people use it.
19:04:34seletzand thus get more ports (and more ports _back_)
19:04:49erikmthat's already happening. the uclinux-dev people started to discover it
19:05:29seletzright now i think that some people think that tey're left alone, see some magic code, get afraid and rather use something different.
19:07:11erikmwhat about this: I'll put the IRC channel in this release, we'll see what happens. if it becomes too crowded, we'll remove it from the documentation
19:07:12seletzthats what happened to me. then i did try bootldr, redbood angel and all that stuff. better i started with blob right away.
19:07:57seletzok, and put in a warning that #blob is for "real" blob talk only.
19:08:05erikmwill do
19:08:07seletzor something like that.
19:08:36seletzremembers 'bout a missing diag docu
19:08:53seletzerikm: when do you plan a release? today?
19:09:08erikm" The documentation is still not in sync with the code" :)
19:09:13erikmseletz: today
19:09:36seletzhmm.
19:10:21seletzjust give me a hour, then i'll ack up some docs for diag.
19:10:34erikmok, put them in doc/
19:11:00erikmseletz: it doesn't have to be much, just tell the people that it should be downloaded and booted as a kernel
19:11:07seletzbut you'll eventually may gtet complaints 'bout bad english (beeing a austrian, that is:)
19:11:29erikmseletz: #include <std-language-disclaimer.h> on top :)
19:11:40seletzlol
19:13:24seletzerikm: btw, do we need a additional mirror for blob?
19:13:38seletzhave some webspace unused
19:13:56erikmI don't think so, though you're free to set up whatever you like
19:22:22akirai hope that 'unused' was not a reference to my work on your site :(
19:22:42erikmakira: which work on whos site?
19:23:10seletzerikm: my site and akiras work :)
19:23:11akiramy work on seletz' site :)
19:23:17seletzdduupp
19:23:40akirafirst thing i'll do after the exams, promised :)
19:24:59akiraright now i'm trying to convince this FLASH Movie her eto pass some variables to included movies, and that bugware gives up on me every 15 minutes.....
19:26:41akira kinda annoying :/
19:51:11seletzerikm: well, diag doc first try done.
19:52:00erikmlooks
19:55:04erikmseletz: good enough. one minor nit: do an s/autor/author/ :)
19:55:19seletzdoh.
19:56:39seletzdone
20:09:27akirahow long will you still be working seletz
20:10:24akira?
20:11:26erikmseletz: oh, I see that I already mentioned #blob in the release notes for 2.0.5-pre1
20:16:21seletzerikm: well, then we didn't got crowded, do we? :)
20:17:41erikmruns "make dist"
20:19:03erikmuntars tarball in /tmp
20:19:47seletzwaiting ...
20:20:55erikmok, it compiles, and a "make dist" tarball contains the same files
20:22:24erikmtags the CVS tree
20:25:27erikmok, the tree is tagged
20:25:40erikmkicks the version number in configure.in
20:26:13erikmgo ahead, the tree is unfrozen
20:43:38seletzerikm: ok
20:46:55BZFlagis the SF project mentioned in that doc? the diff lines included the .nl site...
20:49:57erikmBZFlag: yes, SF is mentioned
20:50:55erikmwaits for the file to show up on SF
20:54:06seletzuff, im tired now.
20:54:23seletzI'll get some sleep, bbl
20:54:51erikmsee ya
20:55:05erikmgoes home after he sends out the release mail
21:00:21seletzgood night :)
22:49:14erikmgoes zzz

Generated by irclog2html.pl by Jeff Waugh - find it at freshmeat.net! Modified by Tim Riker to work with infobot logs, split per channel and by date, etc.