IRC logs for #openrisc Tuesday, 2015-01-13

--- Log opened Tue Jan 13 00:00:53 2015
-!- LoneTech_ is now known as LoneTech08:11
poke53282ysangkok: I am free for both options. Make a suggestion of how to implement it in the json object.08:37
poke53282My initial plan was to include the text config files in the xml file, like /root/.profile08:50
poke53282But that didn't work as expected.08:50
olofk_franck_: Do you know of any limitations in your wb_altera_ddr_wrapper? I'm thinking for example burst lengths and things like that09:30
olofkI'm trying to hook it up to a ddr2 controller09:30
olofkwallento: Hi10:47
wallentohey olofk12:08
wallentoback from my holidays..12:09
_franck__olofk: I don't remember if there is any12:27
olofkwallento: Nice. Did you have a good time?12:48
olofk_franck__: It seems to work for a little while, but then something locks up :/13:01
amswork ..13:04
amsanyone looking for hackers?13:04
olofkI heard Sony were looking for hackers :)13:04
olofkAre you looking for work?13:05
olofkAny particular place?13:05
amsideally, remote .. second place, uppsala, third stockholm region ..13:05
olofkRemote == !Göteborg13:06
amssomething like that13:06
amsdon't mind going to some place once a week, or some such ..13:07
amsgothenburg is close enough13:07
olofkAh, you mean you prefer to work remote13:08
ams:-) what did you think?13:08
olofkI thought you meant remote as in "as far away from Göteborg as possible" :)13:08
amsisn't that .. the same thing?13:09
aburgessams, not really, a cow shed on a hill is pretty remote ... but probably not what you want13:09
amsaburgess: no no, i don't wanna stay in a a cow shed .. i mean gothenburg13:10
olofkI worked from home the whole autumn, which I would count as working remote, but I live and have the office in the same city13:10
amsaburgess: :-)13:11
olofkAnyway, I'll keep my eyes open. Any particular kind of work you're looking for?13:11
amsolofk: well, anything currently .. even short gigs ..13:11
amsolofk: kinda in limbo mentally about what i wanna do13:11
amsolofk: but, Common Lisp, C stuff ...13:12
amsbeen thinking about maybe doing more python, but i hate that language13:13
amsdid python for half the year, and blech ..13:13
olofkThe EDA world is plagued by some unholy alliance with TCL, so for me Python would be a godsend13:18
amsgeda uses scheme13:21
olofkAha. Didn't know that.13:21
amsi can also comb your beard for a nominal fee of a beer a day13:23
stekerncan you do that from remote?13:24
ams... silence, minion!13:27
maxpalnJoining the conversation late, but...I would suggest that anything is possible with a sufficiently complex system of levels and pulleys....14:06
olofkmaxpaln: Solving things with pulleys and levers always felt like doing things like a real engineer14:08
olofkAnd I still haven't gotten around to look at your wb_bfm improvements14:09
maxpaln:-) no problems - they are working well for me. Although as could be predicted my original code didn't handle classic cycles (or constant address cycles for that matter) correctly - it's almost like the actual testing of the code is the process by which you make it work...14:11
maxpalnhmmm, now you've jogged my memory - I realised yesterday the latest version on my dropbox includes some errant debug14:12
maxpalnlet me remove it since you haven't dragged it down yet.14:12
maxpalnhere - use this one when you get to it :-)14:14
olofkI suspect that your DDR controller now might be one of the best tested wishbone peripherals14:17
olofkWhich isn't a compliment to the other peripherals14:18
_franck__olofk: too bad. Do you run a simulation ?14:27
_franck__when I see all this people on the opencores fourm trying to start any openrisc project with or32 and orpsocv2 it makes me think we are not good at all for sharing knowledge14:50
olofk_franck__: Problem is that I don't have any licenses for mixed-language and alteras crappy altmemphy outputs one vhdl file as you know15:10
olofkAnd you're right that we haven't communicated things very well. But another problem is that fusesoc still can't do all that orpsocv2 could15:10
olofkBut we could perhaps make the or1k toolchain work with orpsocv215:11
olofkotoh, I can't build the toolchain with upstream binutils and newlib (Hi wallento), so we have some work to do there as well15:12
olofkHow the fuck is the local_size parameter supposed to work for altmemphy? When I set it to 8, I get 16 words back. If I set it to 4, I get 4 words back15:15
_franck__I did simulation. I think there is a verilog model, neek to check my neek board setup15:18
_franck__I mean a verilog model for the problematic vhd file15:20
olofkAh, now I remember the problem.15:21
olofkYou're right.15:21
olofkBut if I generate a verilog IP instead of VHDL, I get a completely different interface15:21
olofkWhich is just completely insane15:22
olofkWhy Altera? Why!?!?!15:32
maxpalnAH, if only the whole world felt that way :-)16:16
maxpalnon the plus side, once the BFM updates are generally available it should aid the testing of all future peripherals.16:17
_franck__wallento: do you have a source for common systemc models (i2c memory, sdram, nor,...)16:20
stekernolofk: the formula 2^local_size/16 fits your results, have no idea if that match reality though16:34
stekernerr, no it doesn't16:35
stekernOT: can someone explain why åäö doesn't work in irssi over screen on my laptop, but it does in all other windows in the same screen?16:39
stekernand it works from my workstation (and other places as well)16:40
stekerncan't we just deprecate all non-ascii characters, the world would be so much easier then...16:41
stekernblueCmd: I looked at the fetch_and_<op_name>_mask again, but I think you already did the right thing(tm) there16:47
stekernwith l.and   \t%3,%3,%4 # fetch_<op_name>: mask result16:47
blueCmdstekern: \o/16:49
ysangkokpoke53282: how is the file repository made? do you have an automatic script that fetches the packages from original sources and builds them? are you using or1k-debian?18:01
poke53282ysangkok: You linked two days ago to the corresponding scripts.19:17
poke53282No, I am using or1k-macke. and macke stands for "Linux from scratch", "Buildroot", "Sabotage Linux", "alpine linux", "stack overflow", "google" and a few others, which I can't remember right now.19:20
ysangkokoh, it seems i started a naming trend :P19:20
poke53282last link19:21
ysangkokah , yes, thank you19:21
poke53282If I would use debian the filesystem would have hit the 1GB mark half a year ago.19:23
ysangkokpoke53282: even if you only include the packages which are currently included?19:24
poke53282vim takes already 30MB.19:24
poke53282all docs, examples, languages files and so on.19:24
poke53282And do you know, that some fonts need >10MB.19:24
poke53282I think perl needs 80MB19:25
poke53282The same for python.19:25
poke53282all man and info files19:25
poke53282"Glory to 'rm'"19:26
ysangkokheh well, deleting files doesn't  mean that an automated build process is excluded, i guess... it just means certain things must be deleted before packaging19:27
poke53282Yes, but debian also compiles with all possible features.19:27
poke53282configure --enable_ ......19:27
poke53282My distri is built most often with configure --disable.....19:28
poke53282yes, also Debian can be reduced (look at CoreLinux)19:29
ysangkokmy only concern is the lack of a package manager, this is why i am asking...19:31
ysangkokwhat if we faked the installation of core linux packages, so that apt-build would still work, but wouldn't try to build what's already there?19:31
poke53282Yes, I started with my own scripts sith a size of  1-2kB. They grow, and I didn't stop.19:32
poke53282There are tons of other package managers, which are smaller.19:32
poke53282also apt is just the interface as far as I know.19:33
ysangkoki have been researching package managers also, but many of the source-based solutions are too heavy19:33
poke53282I like pacman.19:33
poke53282from archlinux19:33
poke53282emerge seems to heavy, and I think apt too.19:34
ysangkoki never tried it but i believe Arch uses PKGBUILD's in the underlying layer, which requires bash, no?19:34
poke53282But I could be wrong.19:34
ysangkokportage is python and heavy, but emerge/ebuilds are bash19:34
poke53282Yes, arch is using pkgbuild. But pacman is not the tool to build the packages. That is just the manager for binariy packages.19:34
poke53282pacman is C as far as I know.19:35
poke53282We can also look at sabotage and alpine. Those are the distributions based on musl.19:35
poke53282To build a package manager is not a big task.19:36
poke53282to build a good one is difficult.19:36
ysangkokdo you mean build as in compile, or do you mean build as in "design and implement"? :P19:36
poke53282design and implement19:36
ysangkokwell, i am not even considering doing that, that's too much work for me :P19:37
poke53282a few undred lines in bash. and 50 more lines if you want dependencies.19:37
ysangkokyes, but it would be nice to use an existing solution so that we could reuse the PKGBUILD files, for example19:38
poke53282But if you look close, I am using in some way a package manager.19:38
poke53282I have lots of .tar.bz2 files19:38
poke53282and the filesystem is the endproduct of the package manager.19:38
ysangkokyes, but it's on the VM host level, not in the emulator19:38
poke53282buildfs and fs2xml.c are the package managers.19:38
ysangkokyes, i recognize that19:39
ysangkokbut if Max Mustermann goes to and writes irssi, and it doesn't work, it would be nice if he could just build irssi in the emulator using one command...19:40
ysangkokthis is what source-level packages provide, and what is imho missing, and shouldn't be difficult to include19:40
ysangkoki am thinking of things like ebuilds or PKGBUILD19:41
poke53282Unfortunately you miss some part.19:42
ysangkokwhich is ? :)19:42
poke5328250-70% of development time of jor1k work went into the distribution.19:42
poke53282openrisc is not arm and not x86 and not mips. We don't have 100-10000 developers, who work day and night to support their architecture.19:43
poke53282the debian scripts simply don't work.19:43
poke53282We don't have a 100% working glibc version.19:43
poke53282And that's what we would need.19:44
poke53282the Linux kernel still lacks some of the simplest features like a full ptrace implementation.19:44
poke53282I have to cross compile.19:44
poke53282There is no distribution which support cross compiling up to packages like firefox.19:45
poke53282you saw the error yesterday about tmux?19:45
ysangkoki guess you don't think gentoo's/arch's scripts would work any better than debians?19:45
ysangkokbecause of the lacking glibc support?19:46
poke53282I don't know the error yet, but it is defeinitely related to the incomplete toolchain.19:46
poke53282Yes, they won't work.19:46
poke53282And glibc is huge compared to musl and libc, which increases the loading time significantly.19:46
poke53282glibc would be the first step.19:47
poke53282bluecmd was doing this and could compile parts of Debian.19:47
poke53282I wanted to support it, but there is still one bug left, which he needs to fix.19:48
poke53282A chroot works by the way.19:48
poke53282apt works too.19:48
poke53282I tried it some time ago.19:48
poke53282but I don't think, that you want to work with it in the web browser.19:48
poke53282I forgot, but I think that even an chroot with the loading of bash took 1 minute.19:49
ysangkokruntime in jor1k?19:49
poke53282bash compiled with all features, so it loaded also 5 libraries19:50
poke53282yes, emulated in jor1k.19:50
poke53282look at scripts/progs.make19:51
poke53282in the toolchain19:51
poke53282that's basically my distribution19:51
ysangkokah yes19:53
ysangkokthis is what i was looking for19:53
poke53282well, it is not so bad. More and more is working. And at some point will work too, which is the first real real complex program.19:55
poke53282But it took me already 4-6 days just to compile it.19:55
poke53282I am far from running it.19:55
poke53282because I hit several limitations.19:56
ysangkokpoke53282: did anyone isolate the tmux bug? it would be interesting to build a minimal test case19:57
poke53282anyone would be me I guess.19:58
poke53282But that's not the first program, which behaves like this.19:58
poke53282Will be either a problem with the kernel or with the ABI or with musl.19:58
ysangkoki had the same problem with paludis :P20:00
ysangkoka package manager that i built for openrisc but it segfaults when i launch it20:01
ysangkokbut after reading it's configuration files20:01
ysangkoki can strace it20:01
poke53282bould be also a little and big endian problem.20:01
poke53282I had already alls kinds of problem.20:01
poke53282Most of the problems I hit with compiling for openrisc can be related to stupid programmers.20:04
poke53282If the programming world would be perfect we would have ported everything ages ago.20:05
ysangkokheh well you can say that about most businesses :P20:05
poke53282But even the CPU-specific ABI is implemented in around 5-10 packages  Even Firefox has it's own implementation - at two places.20:06
ysangkokthe result of a architecture monoculture :P20:07
poke53282Look at arm64. They did all this stuff within let's say two years. But at least with 30 workers and a lot of help from the community.20:08
poke53282even the JAvascript JIT compilers get support for ARM64.20:08
poke53282and the best thing is, that there is now risc-v and they start from scratch.20:10
poke53282instead of fixing the Openrisc stuff.20:11
ysangkokhmm didn't know about RISC-V20:11
poke53282In short, RISC-V is a fork of Openrisc. I know it is not, but it looks a little bit like it.20:13
poke53282Most of the reasons are not really valid.20:14
poke53282They didn't reach the level of Openrisc yet.20:17
Frans_Hello, any admins online?20:18
Frans_(from opencores)20:18
poke53282Ok, the website is better :)20:19
ysangkoki wonder if they stand to gain anything from making their own ISA20:20
olofkI can understand why someone want to emulate a CPU and run Linux in javascript, but why on earth did Mozilla decide to do a pdf reader in js? It's incredibly slow and buggy compared to evince20:20
olofkpoke53282: Any website is better than OpenCores and the other scattered OpenRISC resource pages20:21
poke53282I guess I can find a website, which is worse.20:22
olofkPerhaps the RISC-V team could have done some web design for us. That's probably a comparable task to doing a new cpu arch :)20:22
poke53282wallento did a website. But I haven't heard anything.20:22
poke53282The first draft looked good.20:22
poke53282I would like to do a virtual machine image with pre-build toolchains.20:23
olofkI looked at it to see if he had any fresh build instructions, but there were just empty links20:23
olofkpoke53282: Yeah. Me too. I don't want to do it, but I want to have it :)20:23
poke53282olofk: I don't understand it either. writing a pdf-reader in JAvascript doesn't make sense.20:24
poke53282But why earth do you think running an Linux emulator in the web browser makes sense.20:24
olofkpoke53282: It doesn't, but the PDF reader is a worse idea :)20:25
olofkAnd not nearly as cool20:25
poke53282I would definitely like to speak with these persons and ask them to go to a psychologist.20:25
poke53282Yep, a pdf doesn't have blinking cursor.20:26
olofkNo. It just has this stupid spinning ring to indicate that it is loading stuff all the time20:26
poke53282You are right. That's less cool.20:28
olofkjor1k is much cooler (but you should still probably go to a psychiatrist)20:28
poke53282after I finished jor1k I have time to do this.20:29
poke53282This is a german parody song on the spinning ring.20:30
poke53282The original song is called the "circle of life" called "Der ewige Kreis" in German, which you can translate to "the permanent circle"20:33
poke53282So the title fits pretty well.20:33
olofkI'll watch it as soon as my poor Atom netbook isn't busy reading PDF files20:35
stekernyeah, even with my limited german skills could figure the humour out even before poke53282 translated ;)20:36
poke53282Kill the windows, which run jor1k. Then it's faster. Be careful, that you not accidently turn off the desktop.20:36
olofkpoke53282: Have you tried to compile a pdf reader for jor1k? :)20:39
poke53282well, firefox compiles.20:39
poke53282olofk: The correct answer is "NOOOOOOOOOOOOOOOOOOOOOOOO.....!!!"20:42
poke53282ysangkok: To come back. Most of the problems I have with compiling stuff are no longer directly related openrisc. I mean, most of the important stuff works as you can see.20:45
olofkhmm.. anyone familiar with avalon-ST interfaces? stekern or _franck_ perhaps?20:46
poke53282The reasons differ.20:47
olofkI'm starting to suspect that the interface exported by wb_altera_ddr_wrapper might actually be an avalon-st interface20:47
poke53282If sabotage linux or alpine linux supports a package it is very likely that it will also work with openrisc.20:47
poke53282ysangkok: In my opinion trying one of those distributions would be the best.20:48
poke53282and with glibc also debian or archlinux is an option.20:49
poke53282But keep in mind, that the reasons for not using them in the end is the simple fact, that they are huge.20:50
poke53282at the moment you download around 3-5MB (jsut guessing) if you want to start jor1k the first time.20:51
ysangkokwell, extra packages don't cost anything except metadata space, which is next to nothing20:52
poke53282This is impossible to reach with any of those distributions.20:52
poke53282at the moment we use musl. And unfortunately musl needs a lot of patches. Because a lot of people use non-standard things from the glibc. And glibc becomes a un-maintainable monster with tons of historic garbage.20:58
ysangkokyes, musl makes sense... but this makes incompatible packages mostly fail on build-time, not run-time, right?20:59
poke53282so in the end we are stuck with sabotage linux and alpine linux.21:02
poke53282because most of the other compile scripts for the packages won't work.21:02
poke53282and don't support cross-compiling.21:03
poke53282but I have to admit, that I didn't compile a lot of games just because they are huge. With huge I mean > 10MB.21:04
olofkThere's a gentoo guy who's built a dist called lilblue, which is uClibc based. Could that be something?21:04
poke53282I cannot ask the users of  jor1k to download so much data.21:04
olofkperhaps not. His goal is to make it hardened. We only care about speed21:05
poke53282Therefore you find mostly game remakes from the old DOS days in jor1k.21:05
poke53282*Ohh, just in case someone missed it*21:06
poke53282the site is awesome.21:07
ysangkoklots of old classics, yeah :D21:09
poke53282the funny thing is, that they also fail in the sound part. no emulator in the web browser get it right.21:11
ysangkokwhat about naclbox? i think the sound is ok there21:15
poke53282this is the chrome native client and does not count.21:16
poke53282Sound is always a problem in an emulation because it depends on the correct timing. In the best case in the millisecond region, in the worst case in the microsecond region.21:18
ysangkokolofk: lilblue is hard but they don't do cross-compilation so it would probably be hard21:18
ysangkokah i meant "is nice ":P21:18
poke53282There are problems on multiple levels. The API of JAvascript, the web audio spoecification and JAvascript itself.21:19
ysangkokpoke53282: did you try out cross-compilation with sabotage linux? their github readme says that they have begun to support it21:19
poke53282All together they make it almost impossible to get sound right in the web browser if you emulate another sound device.21:20
olofkysangkok: Yeah. I remembered wrong. Thought it was a tiny gentoo for MIPS21:20
olofkBut it's more desktop x86_6421:20
poke53282ysangkok: This was my first distribution were I tried it. Unfortunately I didn't try  it with enough patience.21:21
_franck_olofk: what's your problem wi wb_altera_ddr_wrapper ? Don't you have the same interface local_xxx from your altera ddr controller ?21:23
olofk_franck_: Well, first of all it depends on whether I generate a verilog or VHDL IP21:24
olofkI get the local_xxx interface when I generate a VHDL IP, but the documentation for this interface is extremely conflicting21:26
olofkI have read through ~1000 pages of documentation today to find out some more info21:26
olofkThe problem I have is that with the default settings from wb_altera_ddr_wrapper, I seem to get back 16 words when I expect to get 821:27
_franck_I remember I had hard time looking for the right documentation21:27
olofk(using local_size == 8)21:27
olofkSo after reading through all that crap, I'm starting to suspect that it's just an avalon-(ST or MM) Interface.21:28
olofkBut I'm definitely not sure21:28
poke53282ysangkok: You are free to try it.21:28
ysangkokpoke53282: i'll try it later this week when i get some free time again21:29
poke53282musl-cross supports openrisc. That you need as a prerequisite21:29
ysangkokpoke53282: but i would rather fiddle with tmux first21:29
poke53282Well, if you want.21:29
_franck_olofk: isn't that a full rate helf rate config problem ?21:29
poke53282ysangkok: But the problem with tmux is hard. You need a proper toolchain in order to test it.21:30
poke53282to debug it21:30
_franck_you should run a simu with the provided testbench and see what's going on21:30
poke53282Only stekern and me have the proper toolchain.21:30
_franck_I need to reboot if I want to check my ddr settings :(21:30
olofk_franck_: I could, but the verilog version also doesn't provide any testbench21:30
ysangkokpoke53282: i was thinking i'd just remove code until i don't get the segfault anymore21:31
olofkTo make things even better, it looks like they are removing all references to altmemphy and cyclone IV stuff in newer documentation, so they are actively trying to end-of-life it21:31
_franck_fusesoc sim wb_altera_ddr_wrapper doesn't work ?21:31
poke53282ysangkok: And how do you want to compile it? Inside jor1k?21:31
-!- Netsplit *.net <-> *.split quits: clopez, ysangkok, mithro, tariq786, ed-jones, _franck_, enghong, ssvb, stekern, Amadiro, (+27 more, use /NETSPLIT to show all of them)21:32
-!- Netsplit over, joins: hansfbaier, ssvb, felixr, O01eg, ed-jones, fotis, wallento, ams, LoneTech, trevorman (+27 more)21:34
-!- Netsplit *.net <-> *.split quits: clopez, ysangkok, mithro, tariq786, ed-jones, _franck_, enghong, ssvb, stekern, Amadiro, (+25 more, use /NETSPLIT to show all of them)21:36
poke53282ysangkok: I would definitely suggest that you try to compile the toolchain.21:36
poke53282It is not hard. At least, it shoudn't be.21:36
-!- Netsplit over, joins: hansfbaier, ssvb, felixr, O01eg, ed-jones, fotis, wallento, ams, LoneTech, trevorman (+25 more)21:36
-!- Netsplit over, joins: Amadiro21:36
ysangkokpoke53282: musl-cross? i have it already working... but it's using 4.9.0 though, and not 4.9.1 even though stekern says 4.9.1 would work21:37
ysangkokpoke53282: i just used back then... but it's a couple of months ago now21:37
-!- Netsplit *.net <-> *.split quits: tariq786, zama, dalias21:38
ysangkokpoke53282: i used it to compile paludis, which is c++. there is no g++ in the emulator21:38
ysangkokpoke53282: do you think it might crash just because it's c++? d21:39
-!- Netsplit over, joins: tariq786, dalias, zama21:40
-!- Netsplit *.net <-> *.split quits: mithro21:43
-!- Netsplit over, joins: mithro21:43
-!- Netsplit *.net <-> *.split quits: felixr21:43
-!- Netsplit over, joins: felixr21:44
-!- Netsplit *.net <-> *.split quits: fotis, ams, wallento, enghong, mboehnert21:44
-!- Netsplit *.net <-> *.split quits: _franck_, ysangkok, heroux, ssvb, olofk21:48
-!- Netsplit over, joins: fotis, wallento, ams, enghong, mboehnert21:50
poke53282yes 4.9.1 should work. I guess you can change the bash script in musl-cross21:52
poke53282g++ compiled code will work. No problem.21:53
-!- Netsplit *.net <-> *.split quits: knz, ed-jones, blueCmd, atgreen, trevorman, radii21:53
poke53282I didn't put in g++ because .... I didn't.21:53
-!- Netsplit over, joins: ed-jones, trevorman, knz, atgreen, blueCmd, radii21:53
poke53282Therewas no demand from my side.21:54
-!- Netsplit *.net <-> *.split quits: Limb, julzmb, clopez, hno21:54
poke53282But in principle no problem to include it.21:54
-!- Netsplit over, joins: julzmb, Limb, hno, clopez21:54
-!- Netsplit *.net <-> *.split quits: stekern, O01eg, LoneTech, E11fangirl21:55
poke53282Also my or1k-toolchain.tar.bz2 toolchain compiles in some way musl-cross. But with a little bit different options.21:55
poke53282especiall I use the --with-sysroot option.21:55
-!- heroux_ is now known as heroux21:55
poke53282only with this option you can run the rest of my scripts.21:55
poke53282and build the whole stuff for jor1k.21:56
poke53282musl-cross would work, but then I would have to change all my code in scripts/progs.make  and test everything.21:57
poke53282that sounds easier than it is actually.21:57
-!- Netsplit over, joins: stekern21:57
poke53282But then I believe, that you can compiel tmux. You just need ncurses and libevent I guess.21:58
olofk_franck_: That's a good idea22:00
poke53282ysangkok: Do you want to try this?
-!- Netsplit *.net <-> *.split quits: fotis, ams, wallento, enghong, mboehnert22:02
ysangkokpoke53282: if it's necessary. i don't think it will be necessary for tmux, but it might be for sabotage22:03
-!- Netsplit over, joins: fotis, wallento, ams, enghong, mboehnert22:06
poke53282ysangkok: You should first try to compile sabotage for the x86 and take a look in the generated filesystem.22:19
poke53282He has a very special way in handling packages and the whole filesystem.22:20
poke53282the sabotage sabotage developer is a great fan of symbolic links :)22:20
poke53282I like his way of handling the packages, but I don't know if it good.22:22
ysangkokpoke53282: interesting, will do22:26
poke53282well, his whole filesystem is  based on symlinks. There is nothing else22:28
poke53282The files reside in a special folder in which each package has a separate folder.22:29
poke53282The working filesystem then just symlinks.22:29
poke53282The working filesystem contains just symlinks.22:31
ysangkokah reminds me of GoboLinux22:43
ysangkoksounds good22:43
poke53282Yes, reminds me of the goold old DOS days.22:48
poke53282I like neither the design of the Unix filesystem, nor the one of Windows.22:49
poke53282But if I would design my own filesystem I would also end in pure chaos if I would like to support all features of a modern operating system.22:50
poke53282libs, multi-user,  Desktop.22:51
ysangkokyou'd figure it out with enough time and resources :P22:51
poke53282I am not sure if a solution exists. DOS had no libs, no multi-user support and no Desktop. So it was possible. to do it this way.22:53
poke53282One example: I had serveral web servers. One had the website directory in /home/domainname and the other one in /var/www/vhosts/domainname22:56
poke53282which one is better.22:56
poke53282Or should I have one apache directory with the server binaries and in a subfolder the websites. That would be more like the DOS way.22:57
poke53282I don't think, that a good answer exists.22:57
ysangkokdata dependencies are usually configurable as they should... i just wish programs wouldn't use absolute runtime DSO paths22:57
poke53282But I would like to know if you can build a filesystem, that is so logical, that you don't need a manual and is self-explanatory.22:58
ysangkokthere is very little feature-full stuff that you don't need a manual for22:59
ysangkokbut that is ok22:59
poke53282Yes, absolute paths are sometimes terrible.23:00
poke53282You find them too often in codes.23:00
poke53282Have to sleep now.23:01
ysangkokwell the gnu toolchain uses them by default23:01
ysangkokgood night23:01
poke53282yes, the gnu toolchain is a good example.23:01
poke53282good night23:01
--- Log closed Wed Jan 14 00:00:55 2015

Generated by 2.15.2 by Marius Gedminas - find it at!