From: Digestifier To: Linux-Development@senator-bedfellow.mit.edu Reply-To: Linux-Development@senator-bedfellow.mit.edu Date: Wed, 2 Mar 94 03:13:12 EST Subject: Linux-Development Digest #514 Linux-Development Digest #514, Volume #1 Wed, 2 Mar 94 03:13:12 EST Contents: libc-4.4.4 breaks idraw (Dr. Raimund K. Ege) Re: Specialix driver (Robert Sanders) Re: Serial problem with 0.99.15: tty65: input overrun (Rob Janssen) Ytalk binary for linux ( antonio gatta) Re: Linux and X WordPerfect (Rene COUGNENC) Re: serial driver woes (Rene COUGNENC) Re: eth0: transmit timed out in PL15h (Rene COUGNENC) Re: Why not put cluster diffs in nominal kernel before 1.0? (Rick) Re: Help! GCC errors (Dan Miner) Re: undefined symbols in modules (Mike Jagdis) Re: Linux and X WordPerfect (Mike Jagdis) Re: kmalloc fails on 4096 bytes... what is the max now? (Rogier Wolff) Re: kmalloc fails on 4096 bytes... what is the max now? (Rogier Wolff) Re: accessing BIOS Re: structs ip and icmp not defined in pl15h (chianti) NEC T130B scsi host (Francesco Defilippo) Re: _doprnt under Linux... (Johan Myreen) Need to add "kernel build"... (robert myers) ---------------------------------------------------------------------------- From: ege@blitz.fiu.edu (Dr. Raimund K. Ege) Subject: libc-4.4.4 breaks idraw Date: 1 Mar 1994 22:59:44 GMT after upgrading from libc 4.4.4 to libc 4.5.21 my idraw stopped working. It reports Undefined C library functions: 1. light C shared image (Use the real one instead.) when invoked. After removing lic.so.4.5.21 and running ldconfig -v idraw runs again and fine. I understand that the problem lies with the fact that idraw is an old execuatble, produced by an old g++ compiler. Is re-compiling idraw an option ? Does anybody have an idraw execuable that is happy with libc.s0.4.5.21 ? What ever happened to backward-compatibility ? Any help or insight is appreciated. -- Raimund K. Ege School of Computer Science Florida Int'l University ege@scs.fiu.edu (305) 348-3381 University Park ege@servax.bitnet FAX (305) 348-3549 Miami, FL 33199 ------------------------------ From: gt8134b@prism.gatech.EDU (Robert Sanders) Subject: Re: Specialix driver Date: 28 Feb 94 22:43:40 GMT ddj+@cs.cmu.edu (Doug DeJulio) writes: >> Let's say I run Linux, and one of it's system calls is unique to the OS >>(It could be another OS like VSTa for example). If I use that unique system >>call, would my code need to be GPLed under the GPL? I'm not exactly sure if >>Linux has unique (non-public) syscalls, but couldn't that potentially cause >>problems? >That's my understanding, yes. So, you should be *really* careful to >make sure you just use POSIX interfaces. Then there should be no >problem. I don't think this is true at all. I haven't read the GPL that closely, nor have I asked RMS for the FSF's lawyers' opinion on the issue, but the FSF has never stated the Elisp files are automatically GPL'ed, even though they may be run only under Emacs, nor are programs that use GCC extensions like complex numbers or alignment predicates auto-GPL'ed. I think that the current FSF position is more clearly implied by the spirit of the law than by the letter of it. Obviously, writing a Linux program that uses /proc is not creating a "derived work" in the somewhat extended FSF version of the word, but writing a kernel device driver that loads by the 'insmod' modules program at runtime is creating such a derived work. What's the technical difference? Well, they both use Linux-only interfaces (Linux /proc is vastly different from any other existing /proc). Obviously, the latter modifies the kernel, thereby creating a derived work, while the former does not, but I have a very difficult time defending that position. The point is, I think that it is important to make the FSF opinion clear on these matters, even if their opinion might not be legally defensible with the current version of the GPL. Any firm interested in marketing software for Linux should be aware of these issues, not only for their own protection but also so they won't be scared off by incorrect or misleading claims. -- _g, '96 --->>>>>>>>>> gt8134b@prism.gatech.edu <<<<<<<<<--- CompSci ,g_ W@@@W__ |-\ ^ | disclaimer: <---> "Bow before ZOD!" __W@@@W W@@@@**~~~' ro|- erik@errant.north.net (Erik Petersen) writes: >Rob Janssen (rob@pe1chl.ampr.org) wrote: >: > I get this too using 0.99.15(c or e). It has happened when I use >: >mgetty to record a voice file from a ZyXEL modem. I am also set to 56k >: >using a 16550A. What the heck is tty65 anyway, there isn't even one in my >: >/dev directory. >: tty65 = the tty with minor number 65. can have whatever name you choose >: in /dev! >: BTW, this problem does not occur at my system, and I'm using 76800 bps. >: I think it is related to using an IDE disk, which does data transfers >: with interrupts disabled. I'm using an Adaptec 1542B which does not have >: that problem. >Even though this is gettying :-) silly, I'll add my two cents... >I had this problem in a big way and found the problem to involve >using getty on a modem line. Now I'm not sure about exactly what is >going on but...I have three serial ports 2-16450 1-16550A. The modem >is on the 16550A and I have a serial mouse and a 286 running telix >(it has a 16450 so the line runs at 4800) on the other 16450. The >last time I opened my machine, I put the modem on the terminal and >the terminal on the modem designated ports by accident. The next >thing you know, tty65 errors all over the place. >To fix it, I switched the cables and presto, no errors. I think the >problem is in gettydefs and the way the port is initialized. Since >my modem will auto baud the DTE speed (4800 in this case) I don't >think it was overloading the UART. What I suspect (and I'm probably >wrong) is that by using the wrong gettydefs setting the ports went >nuts. I noticed that the modem RXD lights were flashing continuously >(maybe due the the CLOCAL setting) and the terminal didn't work >(fully expected). The errors occured directly parallel with the RXD >burst on the modem. What probably happened: the modem echoed the commands sent to it, and/or replied with OK and ERROR Then, getty was happily chatting with the modem (or with itself, actually). This generates a lot of serial traffic, and thereby exposes the overrun problem which is latently present. uugetty knows how to handle modems, and does not go into this loop. Rob -- ========================================================================= | Rob Janssen | AMPRnet: rob@pe1chl.ampr.org | | e-mail: pe1chl@rabo.nl | AX.25 BBS: PE1CHL@PI8UTR.#UTR.NLD.EU | ========================================================================= ------------------------------ Crossposted-To: comp.os.linux.misc,comp.os.linux.help From: st92ba44@dunx1.ocs.drexel.edu ( antonio gatta) Subject: Ytalk binary for linux Date: Wed, 2 Mar 1994 02:09:06 GMT hello out there. i'm looking for a copy of ytalk (any working version will do) that has been compiled for/on a linux box. actually, even if it hasn't, as long as it *works* on a linux box. please email me anything that would help. thanks in advance. -antonio st92ba44@dunx1.ocs.drexel.edu ------------------------------ From: rene@renux.frmug.fr.net (Rene COUGNENC) Subject: Re: Linux and X WordPerfect Date: 1 Mar 1994 03:10:41 GMT Reply-To: cougnenc@itesec.ensta.fr (Rene COUGNENC) Ce brave Jerry Whelan ecrit: > Other than the size difference, is there any technical reason why > Linux shouldn't just adopt the ELF format as the native binary format? It was the second question I planned to ask, depending on the answer to the first one :-) -- linux linux linux linux -[ cougnenc@renux.frmug.fr.net ]- linux linux linux ------------------------------ From: rene@renux.frmug.fr.net (Rene COUGNENC) Subject: Re: serial driver woes Date: 1 Mar 1994 12:49:05 GMT Reply-To: cougnenc@itesec.ensta.fr (Rene COUGNENC) Ce brave Greg McGary ecrit: > * Is the Linux serial driver generally known to be buggy/unreliable at > high speed, or is there something about my hardware configuration > that's causing me problems that others don't have? I have been using it for months, started using the actual code whith the serial patches against pl14, whith no particular problems. I get an average speed of 1650 cps in v32bis mode, whith a 38400 link to the modem, and whith 16550 or 16450 UART's. (but obviously getting tty overruns whith the 16450). The (386/25) machine does intensive dialouts and receives about 20-40 dialin calls a day, everything is fine. Ftp transfers whith a PPP or SLIP connection give 1.3K / 1.4K speed. I also have sometimes my old MS/DOS PC/XT connected as dumb terminal at 9600 bauds, it works. As you see, I'm an happy user of the serial code :-) -- linux linux linux linux -[ cougnenc@renux.frmug.fr.net ]- linux linux linux ------------------------------ From: rene@renux.frmug.fr.net (Rene COUGNENC) Subject: Re: eth0: transmit timed out in PL15h Date: 1 Mar 1994 12:59:24 GMT Reply-To: cougnenc@itesec.ensta.fr (Rene COUGNENC) Ce brave Peter Suetterlin ecrit: > eth0: transmit timed out, TX status 0x9c, ISR 0x0 > eth0: Possible network cable problem? > 486DX/2-66VL, 3com 503 Etherlink card, Slackware 1.1.0, PL15h ^^^^^^^^^^^^^^^^^^^^^ I have the same card in one of my two machines. When there is a small network problem, like the Ethernet disconnected during less than 1 second, the card seems to hang (not always), and I get these messages. I have to issue a COLD reboot to correct this. No problem whith the Ne2000 clone plugged in my second machine... But most of the time, it works fine. -- linux linux linux linux -[ cougnenc@renux.frmug.fr.net ]- linux linux linux ------------------------------ From: pclink@qus102.qld.tne.oz.au (Rick) Subject: Re: Why not put cluster diffs in nominal kernel before 1.0? Date: Wed, 2 Mar 1994 03:02:58 GMT urlichs@smurf.noris.de (Matthias Urlichs) writes: >Yes. It's a bug in refill_freelist; it doesn't check whether enough buffers >already are on the list. Since refill_freelist is called from read_block >unconditionally, this results in a slowdown or a hangup. >The fix is rather simple. Thanks for the fix. Have you come across a fix for a kernel panic that says "Free list contains shared buffers"? I got this last night - the system was relatively quiet, having finished a fairly large compile (lots of swapping) about 15 minutes earlier. Rick. ------------------------------ From: dminer@nyx10.cs.du.edu (Dan Miner) Subject: Re: Help! GCC errors Date: Wed, 2 Mar 94 04:45:54 GMT In article <2kuitm$or1@dawn.mmm.com>, Dean Junk wrote: > >I just upgraded to .99.15 kernel from .99.14 kernel along with the >following libraries: > > ld.so.1.4.3 > libc.so.4.5.21 > libm.so.4.5.21 > >and the following tar archives: > > image-4.5.21 > inc-4.5.21 > >I am having the following problem compiling xmix: > >/usr/lib/libgcc.sa(__libc.o): Definition of symbol __NEEDS_SHRLIB_libc_4 (multiply defined) >/usr/lib/libc.sa(__libc.o): Definition of symbol __NEEDS_SHRLIB_libc_4 (multiply defined) >make: *** [xmix] Error 1 > >Do you have any ideas? I have everything else working great but this! Yea! Another one bites the dust. :) You didn't notice in the release notes that it tells you to remove/backup/something the libgcc.* files? If you do that, many of your problems will magically disappear. *grin* Dan -- Dan Miner dminer@nyx.cs.du.edu Future student Linux: try it, you'll like. "Your program is encoded in pi." I started with a 64 ------------------------------ From: jaggy@purplet.demon.co.uk (Mike Jagdis) Subject: Re: undefined symbols in modules Date: Tue, 1 Mar 1994 19:50:00 +0000 * In message <1994Feb26.222416.9740@unlv.edu>, Frank Lafaro said: FL> How about someone changing the modules code to not depend on a static FL> ksyms list? If I while having a system running, want to use a module to FL> load code that depends on functions or variables not FL> provided for in ksyms, FL> I have to recompile after adding it to ksyms and reboot FL> before I can use FL> the module. This defeats the purpose of loadable modules. Yeah, I'd noticed that. Still it would be easy to change. When did you want it? Or is someone else already on it? Speak now or go do something more fulfilling with your life... Mike ------------------------------ From: jaggy@purplet.demon.co.uk (Mike Jagdis) Subject: Re: Linux and X WordPerfect Date: Mon, 28 Feb 1994 23:06:00 +0000 * In message , Thomas Boutell said: TB> I encourage those TB> who want to know why commercial product XYZ isn't available TB> for Linux to instead ask why Linux doesn't yet fully TB> support COFF, ELF, IBCS, etc. It does. It's alpha but it does. Or, more specifically, we can load COFF format binaries, we can load ELF format binaries and we can emulate both SVR4 (popularly refered to as iBCS2 compatibility by commercial vendors with SVR4 based offerings) and SVR3 kernel interfaces - SVR3 in, at least, SCO, ISC and Wyse Unix V/386 flavours including TCP/IP. Even more specifically I have the Uniplex office automation suite, Informix SQL/SE and WordPerfect 5.1 (all SCO versions) running happily. WordPerfect works in X mode as well when talking to the display via TCP/IP at localhost:0. TB> (There's work being done in this area, but I'd like to TB> hear more from those doing it and I'd like to hear TB> how I might be of help. This is a very important TB> area for Linux.) Join the IBCS2 channel on the linux-activists mailing list. Right now you can write tests - lots of tests :-). Everything from short programs that exercise a single system call - return values, boundary conditions, over/under runs, bad data handling - to great monolithic stonkers! It's easier to debug from code than to employ psychic abilities on real applications :-). Mike ------------------------------ From: wolff@hal.et.tudelft.nl (Rogier Wolff) Subject: Re: kmalloc fails on 4096 bytes... what is the max now? Date: 1 Mar 1994 22:10:35 GMT jmv (jmv@Rabbit.edu) wrote: : - Unlike what I used to believe, because of the header, optimal kmalloc : request are not powers of 2. : For example, as I understand from the get_order function in linux/mm/kmalloc.c , : when you ask for 1024 bytes, it's going to get 2048, because the available : data in a 1024 block is only 1020. Am I wrong ? Nope you're almost right..... You lose 16 bytes per "page". Next you loose 8 or so bytes per block that would reduce your 1024 block to ((4096-16)/4 ) - 8 = 1020-8 = 1012. I just checked the source: I'm right..... The "optimal" values can be calculated from the table in kmalloc.c by subtracting 8 this leads to the following numbers: 24, 56, 120, 244, 500, 1012, 2032, 4072. The 8 is found in the "size += sizeof (struct block_header);" in get_order. Nice to see that others can read my code, and figure out what is going on..... Roger (author of current kmalloc) Wolff -- * Not that I have tested it - I just wrote the code and hope it works. * * "Real programmers" don't test: they assume it works the first time, * * and anyway, what do you think beta-testers are for? -Linus Torvalds * EMail: wolff@dutecai.et.tudelft.nl ** Tel +31-15-783643 or +31-15-142371 ------------------------------ From: wolff@hal.et.tudelft.nl (Rogier Wolff) Subject: Re: kmalloc fails on 4096 bytes... what is the max now? Date: 1 Mar 1994 22:29:44 GMT Theodore Ts'o (tytso@ATHENA.MIT.EDU) wrote: : From: jmv@Rabbit.edu (jmv) : Date: 25 Feb 1994 05:24:28 GMT : The max kmalloc you can do is 4080. Why ? because there is a 16 bytes : header. As far as I could tell (tell me if I am wrong), this is only defined : This is only true if you have the Kernel debugging malloc option turned : on. If it is off, then there is no 16 byte header, and you can malloc : up to 4096 bytes. Not true. I should know: I wrote it. (You wrote the old one I know....) (How things can change if you don't read the sourcecode for a few months....) Roger. -- * Not that I have tested it - I just wrote the code and hope it works. * * "Real programmers" don't test: they assume it works the first time, * * and anyway, what do you think beta-testers are for? -Linus Torvalds * EMail: wolff@dutecai.et.tudelft.nl ** Tel +31-15-783643 or +31-15-142371 ------------------------------ From: whitney@galileo.Meakins.McGill.CA () Subject: Re: accessing BIOS Date: Tue, 1 Mar 1994 23:14:35 GMT Robert Sanders (gt8134b@prism.gatech.EDU) wrote: : whitney@galileo.Meakins.McGill.CA writes: : >: >So, the question remains : How do I get from protected mode : >: >to real mode and back ? : >I am not intrested in the v86 feature for my experiment - I want : >to know how to switch to real mode. : If you switch to real mode, you'll kill Linux, plain and simple. : Real mode uses a different interrupt vector scheme, so all the : Linux interrupt and trap handlers will be useless. I didn't see : the original post, but if you really want to go into real : mode [snip] Thanks for the comments, but I can see I posed my question to the wrong group. My intrest was in switching mode on the '386 so that I could easily move between BIOS based code and a 32-bit flat memory space - ie. no multitasking, or paging. For my needs Linux, NT, OS/2, DOS extenders, etc only complicated matters. In fact switching between modes requires only a page of code. Whitney ------------------------------ From: chianti@dchia.gse.rmit.edu.au (chianti) Subject: Re: structs ip and icmp not defined in pl15h Date: Tue, 1 Mar 1994 23:18:09 GMT In article jrozes@allegro.cs.tufts.edu (J Rozes) writes: >I was just porting the NetBSD timed and interestingly, I found that > and do not define struct ip and struct icmp, >respectively. I did some grep'ing to see if they were defined elsewhere, >but came up with nothing. > Dont know about pl15. In pl14 I recalled linux has different struct's than those in BSD. Even those defines like ICMP_ECHO are spelled differently. An attempt to avoid copyright suit? :) ZZZzzz ------------------------------ From: clint@hal9000.unipv.it (Francesco Defilippo ) Subject: NEC T130B scsi host Date: 2 Mar 1994 07:48:13 GMT Hello, The Nec T130B scsi is supported by linux? -- With Best Regards: +--------------------------------+ | Francesco Defilippo | | clint@hal9000.unipv.it | | pk: finger(1) e-mail | +--------------------------------+ +--- Net .... ^ ^ / 0 0 / =--------------oOO-(_)-OOo--------------------= ..attenti qualcuno vi osserva.. ------------------------------ From: jem@snakemail.hut.fi (Johan Myreen) Crossposted-To: comp.os.linux.help Subject: Re: _doprnt under Linux... Date: 02 Mar 1994 07:58:49 GMT In article <2l0nrh$435@darkstar.UCSC.EDU> galpin@cats.ucsc.edu (Dan) writes: >In porting an app to Linux, I discovered that Linux does not seem to >have the "standard" UNIX _doprnt library function.. (this is with gcc). >Is there an equivalent function that I could use? "Standard"... My personal opinion is that anybody using _doprnt deserves to be punished. It is not documented and is not intended to be used. It is callable (on sytems where it is available) because of a flaw in the data abstraction mechanism. See the vprintf, vsprintf and vfprintf functions. -- Johan Myréen jem@cs.hut.fi 60° 11' 55" N, 24° 53' 30" E ------------------------------ Crossposted-To: comp.os.linux.help From: bmyers@netcom.com (robert myers) Subject: Need to add "kernel build"... Date: Wed, 2 Mar 1994 07:41:55 GMT I've just tried to get a compiled .99p14d kernel installed on my home linux system (slackware 1.1.0 based). Apparently, due to my not having compiled a kernel and installed directly on a hard disk (instead of using a boot floppy), I somehow corrupted my system's ability to boot off my esdi hard drive. I've found an old posting in c.o.l.a. about getting a bootfloppy from the origional installation and running /etc/lilo/config from it (I'll be doing that later today, when I bring my 1.1.0 disks from work home). However, i think that a few things need to be changed/added: 1) the basic HOWTOs should say how to compile and install a new kernel to boot off of floppy, hard drive or whatever. examples using lilo should also be given. 2) need to change README file in kernel source distribution to go through specific steps on getting the compiled image to the right place (not assume that an experienced person is doing this). I'll be sending a message later today to Linus requesting this. -bob ------------------------------ ** FOR YOUR REFERENCE ** The service address, to which questions about the list itself and requests to be added to or deleted from it should be directed, is: Internet: Linux-Development-Request@NEWS-DIGESTS.MIT.EDU You can send mail to the entire list (and comp.os.linux.development) via: Internet: Linux-Development@NEWS-DIGESTS.MIT.EDU Linux may be obtained via one of these FTP sites: nic.funet.fi pub/OS/Linux tsx-11.mit.edu pub/linux sunsite.unc.edu pub/Linux End of Linux-Development Digest ******************************