From: Digestifier To: Linux-Development@senator-bedfellow.mit.edu Reply-To: Linux-Development@senator-bedfellow.mit.edu Date: Sun, 11 Sep 94 16:13:09 EDT Subject: Linux-Development Digest #157 Linux-Development Digest #157, Volume #2 Sun, 11 Sep 94 16:13:09 EDT Contents: S3 and 24bit or 16bit Curses help (Benjamin B. Rickett) Another DOSEMU.52 Problem. . . (Grungie The Wise) Re: S3 and 24bit or 16bit (Mihail S. Iotov) Re: Alpha Linux (Johan Myreen) Re: Why I cannot mount a PhotoCD on Mitsumi ? (Jeff Kesselman) File locking--gurus please read. :) (Willis Boyce) Re: Survey: who wants f77,cc,c++,hpf for linux? (Mike Castle) Re: 320x200 X resolution? (Christopher M. May) Re: floppy problems in 1.1.49 (Olav Kvittem) HP Ethercards Not Supported (Michael Talbot-Wilson) DOOM for Linux problem - help. (Dane Jasper) Re: Survey: who wants f77,cc,c++,hpf for linux? (Robert Sanders) Re: Wheres blkdev.h?? (compiling 1.1.49) (Stephen Harris) 480x360 Res works for me. (Daniel L Moore) Re: Survey: who wants f77,cc,c++,hpf for linux? (Carlos) Re: Developing Distributed Filesystems for Linux? (Derek Atkins) Re: Developing Distributed Filesystems for Linux? (John F Carr) Re: Alpha Linux (Richard Henderson) ---------------------------------------------------------------------------- From: hwj@henrik.igk.dth.dk () Subject: S3 and 24bit or 16bit Date: 11 Sep 1994 17:08:33 GMT Are there any software on the net that makes it possible to show 24 bit graphics or perhaps only 16 bit on S3-based graphics cards (including 805). Svgalib only describes a faulty/non completed S3-driver. The software/driver only has to be able to plot 24bit-pixels. I know that XFree86 v3.1 should have 64k-colours on the S3 but I would like to have solved this problem a little earlier... - Henrik Wann Jensen ------------------------------ From: rickett@highway.alinc.com (Benjamin B. Rickett) Subject: Curses help Date: 11 Sep 1994 04:21:49 GMT I am modifying my curses programs to ahve color, but i'm not exactly sure where to start. Does any body have a simple example of how to use color in curses? If youu do please let me know. Thanks... Please send Email to SLKWR@CC.USU.EDU -- Internet Alliance Administrator -Benjamin B. Rickett ------------------------------ From: slash@cyclone.Stanford.EDU (Grungie The Wise) Crossposted-To: comp.os.linux.help Subject: Another DOSEMU.52 Problem. . . Date: 11 Sep 1994 10:23:57 -0700 When compiling Dosemu0.52, I get the following error: make[1]: Entering directory `/usr/src/dosemu0.52/mouse' gcc -O2 -Wall -I/usr/src/dosemu0.52/include -I/usr/src/dosemu0.52 -I/usr/src/linux/include -c mouse.c mouse.c: In function `fake_int': mouse.c:587: `VIF_MASK' undeclared (first use this function) mouse.c:587: (Each undeclared identifier is reported only once mouse.c:587: for each function it appears in.) mouse.c: In function `mouse_do_cur': mouse.c:672: warning: implicit declaration of function `mmap' mouse.c: At top level: mouse.c:156: warning: `mousescreenmask' defined but not used make[1]: *** [mouse.o] Error 1 make[1]: Leaving directory `/usr/src/dosemu0.52/mouse' make: *** [dossubdirs] Error 1 I commented out the bad code, and it compiled, but I got the same errors again in dpmi.c. Thats as far as I got. I dont have time right now to search the code myself, so I was wondering if anyone has had similar problems, or knows what VIF_MASK is and where is is defined. (It has to be a macro expansion, of LWORD, I think, as it is not directly in the code, but I didnt have a chance to run it through the preprocessor) Any help or comments would be appreciated. Hopefully I might find some time to figure it out. Jeff -- __________------== Jeff Townsend ==------___________ ____----DCC Consultant - Guitarist - CS Major - Simpsons Fan----____ -==slash@cyclone.stanford.edu jefft@xenon god@cs grungie@leland ==- ----_______"Yes sir. Very much so sir. Obviously insane."_______---- ------------------------------ From: iotov@cco.caltech.edu (Mihail S. Iotov) Subject: Re: S3 and 24bit or 16bit Date: 11 Sep 1994 17:38:26 GMT hwj@henrik.igk.dth.dk () writes: >Are there any software on the net that makes it possible to show 24 bit >graphics or perhaps only 16 bit on S3-based graphics cards (including 805). >Svgalib only describes a faulty/non completed S3-driver. >The software/driver only has to be able to plot 24bit-pixels. >I know that XFree86 v3.1 should have 64k-colours on the S3 but I >would like to have solved this problem a little earlier... Try some DOS program under DOSEMU. My S3 does not work with these either, but you may have more luck. ------------------------------ From: jem@snakemail.hut.fi (Johan Myreen) Subject: Re: Alpha Linux Date: 11 Sep 1994 17:34:23 GMT In article <779290449snz@lepton.demon.co.uk> nick@lepton.demon.co.uk (N J Plant) writes: >On the 68000 the external address bus is 20 bits and the external data bus >is 8. The external data bus is 16 bits wide on the 68000. > Internally, the registers, buses and ALU are all 32 bit. It can ADD >and SUBtract 32 bit numbers or MULtiply 2 16 bit numbers to give a 32 bit >result. It has fewer pins than a 68040, but its still a 32 bit chip. The >sizeof the integral types should be the same as any other 68K chip. 32 bits? Do you really think that is the natural integer size considering the 16 bit external data path and the lack of a 32x32 bit multiplication instruction? According to the C standard both char and short operands in an expression are promoted to int before the operation proceeds. This means all arithmetic operations would be performed with 32 bit operands, which is *not* *too* *efficient* if all you've got is a 16x16 multiplication instruction. (Well, I guess a compiler can optimize sometimes, but not always.) >The Intel 8088 has a 20 bit address bus, but is otherwise an 8 bit device. >So: > short = 8 bits > int = 16 bits Hmm, I thought you said it was an 8 bit device? > long = 16 bits >Neither of these processors has the address space or MMU necessary for >Linux to run on them, so the natural word size is irrelevant. I guess you mean irrelevant to this newsgroup. But there are C compilers for these processors too. -- Johan Myreen jem@vipunen.hut.fi 60 11' 55" N, 24 53' 30" E ------------------------------ Crossposted-To: comp.os.linux.help From: jeffpk@netcom.com (Jeff Kesselman) Subject: Re: Why I cannot mount a PhotoCD on Mitsumi ? Date: Sun, 11 Sep 1994 17:26:22 GMT In article <34to9k$dit@rutcor.rutgers.edu> badics@rutcor.rutgers.edu (Tamas Badics) writes: >jeffpk@netcom.com (Jeff Kesselman) writes: > >[many interesting technicalities deleted...] > >>The poitn of all this is that ALL CD-ROM types are fully ISO9660 >>compatable. (other than perhapse really wierd propritary formats-- I won't >>even guess what the Atari Jaguar does). > >>jeffk@crystald.com > >OK, it is all nice to know, but how can I read a PhotoCD on a Mitsumi drive? >(My MS-DOS driver can read them without problems.) > >Tamas > > Boy, after all that info I put out, I have to admit that THIS question I'm not 100% sure of... mostly 'cause I've never had to do it (all my CD-ROM programming has been on dedicated platforms such as C-I, 3DO, etc...) Why don't you tell us exactly what you are doing, and where it is failing? Are you mounting the CD-ROM using Linux mount? Does it mount successfully or error out? We may get into details that will need the trained eye of someone who actually knows the innards of the CD-ROM drivers under Linux but, technically and according to the standards, you SHOULD be able to mount a Photo-CD and see its files. CD type discs can have multiple 'tracks'. On a yellow CD-ROM, the data is all in track 0. I assume Photo-CD by definition would HAVE to be the same... ------------------------------ From: wboyce@panix.com (Willis Boyce) Subject: File locking--gurus please read. :) Date: 11 Sep 1994 13:55:43 -0400 I've been working recently on a DBMS project under Linux. My ultimate goal is to create a DBMS which compares favorably with commercial systems and which provides a very simple interface into C. The primary DBMS I use now (at work) is Sybase, and I am frustrated with the difficulty of interfacing that DBMS to C as well as with the temptation to code a lot of processing in stored procedures. My DBMS will deliberately omit stored procedures and encourage the user to do most processing in C. After looking into various methods (I'm pretty new to Unix), I decided to use advisory file locks as my concurrency mechanism. These have two big advantages: 1. They allow me to use a decentralized approach to concurrency. 2. They have built-in deadlock detection. Unfortunately, the Linux 1.1.8 that I am running apparently doesn't support deadlock detection. Indeed, there is even a comment in flock.c to that effect. Anybody who has ever used a heavily-loaded DBMS knows that deadlock detection is not something that can be done without; therefore, I would like to add it. First, though, I am wondering if anybody has *already* added deadlock detection to Linux. If so, in which version of the kernel did it make its debut? And, what is the most stable kernel version *after* the deadlock detection was added? :) If deadlock detection is still in need of address, my analysis is that it won't be easy. I'm soliciting suggestions on the best way to proceed. I'm interested in the opinions of the other Linux developers, since I would like to contribute my patch if/when it is finished. A simple algorithm that, before blocking task A on B, checks that B isn't already blocked on A won't cut it, because true deadlock detection requires handling a block chain that could conceivably include every process in the system. Given a file_lock structure that is about to be created (called file_lock in the locks.c), and a file_lock structure called fl that file_lock would conflict with, I need to find out if fl's owner is ultimately blocked on a file_lock structure which is owned by the current task, called current. (The quick and dirty way would be to declare a deadlock if fl's owner was blocked at all. That wouldn't be great, but it would work. Opinions?) One problem is that the association between file_lock and wait_queue is only one-way. Given a wait_queue, the only way to get the first file_lock of the waiting task is to search the file_lock chain for a file_lock structure whose owner is the waiting task. It seems like the only to handle it would be a recursive function together with a list of already-processed tasks. The list hold pointers to those tasks which had already been determined to *not* be waiting for current. In the event that A and B were both blocked on C, and C had been determined to not be waiting for current, the list would prevent C from being evaluated a second time. In the event of a deadlock, do I blow away the task's other file locks, or do I let the calling process do it? (Obviously it has to be done, else the process will never be able to obtain its locks.) Anyway, feel free to mail me at wboyce@panix.com or talk-page me at wboyce@wboyce.dialup.access.net (my SLIP site, non-mailable) if anybody feels like discussing this. Or post a reply. :) Will ------------------------------ Crossposted-To: comp.lang.fortran From: mcastle@umr.edu (Mike Castle) Subject: Re: Survey: who wants f77,cc,c++,hpf for linux? Date: Sun, 11 Sep 1994 17:31:04 GMT In article , Dan Pop wrote: >Could you post some examples where a commercial native compiler for x86 >produces _significantly_ faster codes than the free gcc? IBM's C compilers under OS/2. Watcom compilers under OS/2 and DOS (do they have unix versions?) Zortech C compiler (I presume under OS/2 and DOS as well). Most likely the Mark Williams C compiler (they produce better compilers than operating systems (coherent)). GCC is designed to be PORTABLE first, optimal last. In almost all cases, a DECENT architecture specific compiler will be as good as or beat GCC. -- Mike Castle .-=NEXUS=-. Life is like a clock: You can work constantly mcastle@cs.umr.edu and be right all the time, or not work at all mcastle@umr.edu and be right at least twice a day. -- mrc We are all of us living in the shadow of Manhattan. -- Watchmen ------------------------------ From: cmay@titan.ucs.umass.edu (Christopher M. May) Crossposted-To: comp.os.linux.misc Subject: Re: 320x200 X resolution? Date: 11 Sep 1994 17:51:33 GMT Andreas Matthias (andy@titan.central.de) wrote: : : So.... 320x200 resolution anyone? :) : : I'll post if I figure out how to do it. : I have one that's working here (ET4000 with 17'' AOC monitor), but it : occupies only about half of the screen in vertical direction. I did : not find out how to make it bigger vertically. Perhaps someone else : can continue with this: : ********************************************************************** : ModeDB : # name clock horizontal timing vertical timing flags : "320x200" 25 320 360 424 440 200 200 240 250 : ********************************************************************** : btw: Could it be that fvwm gets confused with this resolution? It : seems not to scroll correctly in the vertical direction. : Ciao, : Andreas : -- : Andreas Matthias : Zehntenstr.9 : D-37120 Bovenden : Voice: +49/551/81377 You need a lower dot clock frequency in my opinion. I had the same problem with higher dot clocks. I used a 12Mhz dot clock frequency in the modedb line I posted earlier. 320x240 12 320 352 392 424 240 243 247 252 -- -Chris May, Computer Science, University of MA, Amherst - Technical Assistant, P.C. Maintenance Lab ------------------------------ From: oak@domen.uninett.no (Olav Kvittem) Subject: Re: floppy problems in 1.1.49 Date: 11 Sep 1994 18:32:50 GMT In article <34n8h0$cl9@sunserver.lrz-muenchen.de> u7y22ab@sun2.lrz-muenchen.de (Wolfram Gloger) writes: >: Sep 2 04:58:18 darkstar kernel: VFS: Disk change detected on device 2/0 >: Sep 2 04:58:22 darkstar kernel: floppy: disk absent or changed during operationSep 2 04:58:22 darkstar kernel: floppy I/O error >: Sep 2 04:58:22 darkstar kernel: dev 0200, sector 38 >[ ... more stuff deleted ... ] As this was not while loading a ramdisk, I can't comment. >You're not the only one having problems, Mike. >I get messages similar to the above when I boot from floppy and >the kernel is copying the floppy to the ramdisk. This occurs >with both 1.1.23 and 1.1.45 but not 1.0 or 1.0.9. I had this too until 1.1.47 (I believe) then it got fixed - I have a bootable rootdisk with 1.1.49 that works just fine. I have problems even with 1.1.50, 1.1.37 and 1.1.23 but not with 1.0. I am trying to make my own Slackware bootdisk with a kernel that can do NFS over a D-link adapter. Olav -- Olav Kvittem : UNINETT A/S RFC Address Olav.Kvittem@uninett.no OR Address C=no;ADMD=" ";PRMD=uninett; O=uninett;S=Kvittem;G=Olav Postal Address Box 6883 N-7002 Trondheim Phone +47-7-596981 +47-7-596450(FAX) Description Network Manager "Networking needs neat working - Nettverk er nette verk !" ------------------------------ From: mike@gumleaf.apana.org.au (Michael Talbot-Wilson) Subject: HP Ethercards Not Supported Date: Sun, 11 Sep 1994 13:26:38 GMT The Ethernet HOWTO describes the HP 27247B and 27252A in these terms: "These cards are high performers (3c509 speed) without the interrupt latency problems (32K onboard RAM for TX and RX packet buffering). They both offer LAN connector autosense, data I/O in I/O space (simpler) or memory-mapped (faster), and soft configuration. 27247B was rated Best for ISA Servers by PC Mag this year." This discussion is under the "Supported:" sub-heading. In my hands, however, the kernel does not detect these cards. I have tried the 27247B in a 386SX IDE and the 27252A in a 486DX2 SCSI. (Yes, I know the disk makes no difference.) And I have tried several generations of the kernel, all compiled with HP PCLAN support enabled. I have tried passing the IRQ and base port address in an ether string at the lilo: prompt and I have tried the lilo reserve ... ether... parameters. I compiled a kernel with both HP PCLAN and Intel EtherExpress support. When I booted this kernel with the HP 27252A installed the kernel announced detection of the EtherExpress, and on the same line, detection of an error and rejection of the card. It did not detect the HP card. I booted this kernel with an EtherExpress card installed. That was successful, and for the first time ever the ifconfigs in rc.inet1 ran without error. The kernels I tried were 1.0.8, 1.1.38 and 1.1.49. The HP cards test OK with HP's software under MSDOS and they are detected by Microsoft Windows for Workgroups. -- Michael Talbot-Wilson ------------------------------ From: dane@nermal.santarosa.edu (Dane Jasper) Crossposted-To: alt.games.doom,comp.os.linux.help Subject: DOOM for Linux problem - help. Date: 11 Sep 1994 18:31:00 GMT I am having a problem getting DOOM for Linux to run on one of my machines. On the 486/66 machines at school, things work just fine. It's very depressing not to be able to play DOOM at home! Here's what I get (system info follows): # linuxxdoom linuxxdoom: using incompatible library '/usr/X386/lib/libXt.so.3.0.1' Desire minor version >= 1 and found 0 linuxxdoom: using incompatible library '/usr/X386/lib/libX11.so.3.0.1' Desire minor version >= 1 and found 0 DOOM System Startup v1.666 V_Init: allocate screens. M_LoadDefaults: Load system defaults. Z_Init: Init zone memory allocation daemon. W_Init: Init WADfiles. adding ./doom1.wad shareware version. M_Init: Init miscellaneous info. R_Init: Init DOOM refresh daemon - [...................] P_Init: Init Playloop state. I_Init: Setting up machine state. Could not start sound server [sndserver] X Error of failed request: BadWindow (invalid Window parameter) Major opcode of failed request: 132 (XTEST) Minor opcode of failed request: 1 (X_XTestCompareCursor) Resource id in failed request: 0xbffff6a0 Serial number of failed request: 4 Current serial number in output stream: 4 The system is a 486/66 with 28 megs of ram and linux 1.0.9. X is version 2.1, with the XF_SVGA server. Does anyone have any ideas? Is it the libs?? Replies via email are appreciated - post as well if you think it would be of general interest. Dane ------------------------------ From: rsanders@mindspring.com (Robert Sanders) Crossposted-To: comp.lang.fortran Subject: Re: Survey: who wants f77,cc,c++,hpf for linux? Date: 11 Sep 1994 18:30:04 GMT On Sun, 11 Sep 1994 17:31:04 GMT, mcastle@umr.edu (Mike Castle) said: > In article , Dan Pop wrote: >> Could you post some examples where a commercial native compiler for x86 >> produces _significantly_ faster codes than the free gcc? > IBM's C compilers under OS/2. > Watcom compilers under OS/2 and DOS (do they have unix versions?) > Zortech C compiler (I presume under OS/2 and DOS as well). > Most likely the Mark Williams C compiler (they produce better > compilers than operating systems (coherent)). I think Dan meant he wanted to see the C source, timings, and the assembly produced by each, as well as other relevant benchmark information (input data, RSS, host machine, etc.). I know that at least one of your statements (the last) is untrue; the big benefit of the MWC compiler is that it compiles fast; when Coherent users want code that runs fast, they use GCC. > GCC is designed to be PORTABLE first, optimal last. In almost > all cases, a DECENT architecture specific compiler will be as > good as or beat GCC. This is simply a restatement of the original claim. We're trying to verify it empirically, not take a vote. -- Robert ------------------------------ Crossposted-To: comp.os.linux.help From: hsw1@papa.attmail.com (Stephen Harris) Subject: Re: Wheres blkdev.h?? (compiling 1.1.49) Date: Sun, 11 Sep 1994 10:32:32 GMT Dougal Campbell (dougal@vespucci.iquest.com) wrote: : I ran across the same thing when I compiled the 1.1.49 kernel. The : patches seem to not place some of the files correctly. If you look in The patches are perfect. I have patched from 1.0.0 up to 1.1.50 without any errors. There are a number of things to keep in mind when patching 1) Always keep a CLEAN tree (ie unconfigued etc) around. I generally keep a 'tar' of the sources just before I configure/compile. This is because configure changes files, and so patch might fail. 2) Always path from the /usr/src directory 3) Always use the -p0 flag when patching This is how I do my patching: cd /usr/src mv linux linux.old tar xvf ARCHIVE/linux.49.tar gzip -dc < ARCHIVE/patches/patch50.gz | patch -s -p0 # see if any .rej files exist - shouldn't! find linux -name '*.rej' -print # Remove the old files find linux -name '*.orig' -exec rm {} \; # Make a new archive tar cvf ARCHIVE/linux.50.tar linux I can now configure/compile etc. When the next patch comes along the new tar file I created is a clean source tree to patch against. If you don't understand the above, then you probably shouldn't be playing around with ALPHA level kernels. -- rgds Stephen ------------------------------ From: mooredan@uxa.cso.uiuc.edu (Daniel L Moore ) Crossposted-To: comp.os.linux.misc Subject: 480x360 Res works for me. Date: 10 Sep 1994 01:55:57 GMT Here's a Xconfig line that works for my CrystalScan 1572FS monitor. "480x360" 25 480 496 504 664 360 360 368 377 It'll will probably work with other monitors that have the following specs: Horizontal Scan Freq: 30 - 64 kHz Vertical Scan Freq: 50 - 100 Hz Bandwidth: 80 MHz Also note that the dot clock is 25MHz for your video card. DOOM runs fine, now to find solutions to the ctrl, alt - arrow keys combinations, (my window manager takes over), and getting the sound to work, probably need to upgrade my driver. -- ******************************************************************************* * Daniel L. Moore mooredan@uiuc.edu * * University of Illinois at Urbana-Champaign -- College of Engineering * ******************************************************************************* ------------------------------ From: carlos@posseidon.if.usp.br (Carlos) Crossposted-To: comp.lang.fortran Subject: Re: Survey: who wants f77,cc,c++,hpf for linux? Date: 10 Sep 1994 00:30:04 GMT In article <34m769$bju@indy.pgroup.com> lfm@pgroup.com (Larry Meadows) writes: 1. Are people interested in a commercial compiler suite for Linux on Intel Architecture platforms? The suite would include true compilers for extended Fortran 77, ANSI C, Draft-ANSI C++ with extensions, and High Performance Fortran. C, f77, and C++ could support shared memory parallelism (thread-based) if system support is available in Linux. HPF would support socket-based communications on networked systems, and could support custom interconnects. About C-C++: if your product is better than gcc, yes. By better I mean not being buggy and producing faster code (at least 15%, on a 486, on a wide variety of codes). It doesn't need to produce a lot of warnings, etc. since this can be done with gcc. For FORTRAN, only if it's full f90. Of course it should produce SIGNIFICANTLY faster code than f2c-gcc. 2. How much would people pay for such a product [ loaded question ]? Around $500 is reasonable for one language, with an extra one around $200. 3. What distribution media would be required? ftp, floppy. 4. Is there interest in accompanying GUI/non-GUI debuggers and performance analysis tools? If it's not expensive, otherwise not. Carlos ------------------------------ From: warlord@MIT.EDU (Derek Atkins) Crossposted-To: alt.filesystems.afs Subject: Re: Developing Distributed Filesystems for Linux? Date: 11 Sep 1994 19:57:26 GMT In article <34ou3c$n41@agate.berkeley.edu> lim@soda.CSUA.Berkeley.EDU (Lincoln Myers) writes: Is anyone working on a filesystem for Linux or another freely available UN*X which will offer the advantages that these do? AFS was a project at CMU before it became a commercial product supported by Transarc. Would it be feasable to port an earlier version (pre-3.0?) to Linux? (Is it freely available? Would it be compatible with current AFS? Would one want to use it?) If not, would it be possible to make a freely available implementation of AFS or DFS for Linux, without infringing on their current owner's (Transarc's) rights? Is there enough information out there? Well, I can try to answer _some_ of these questions. I have ported Transarc's AFS version 3.3 to Linux. It is currently in Beta testing at MIT and soon to be in testing at a few other AFS Licensee sites. We are _hoping_ to get the rights from Transarc to distribute binaries to all comers, but at this point in time we are not in a position to do so. The distribution is only in object form (no sources), and it comes as a loadable kernel module (so no, the GPL does not, and need not, apply, so distribution is not illegal). No changes are required to the kernel in order to run Linux-AFS, although it does require a Linux kernel version of 1.1.33 or higher, and unfortunately because of the way modules are supported in Linux, it might require a recompile if certain things change in the kernel. Right now, the beta test is using 1.1.48 and 49, and I believe it will also work with 1.1.50 without a recompile. The current testing and usage shows it to work fairly well under light load, although there are problems with high usage. For example, linking "gdb" into AFS appears not to work (although I haven't tried to reproduce this, yet, so I haven't verified this behavior). So, to sum up the current situation: Linux-AFS exists. It is in beta-test. You currently need a Transarc license to use it, and we need to get permission for each site that wants to get a copy of it. (We currently have permission for ISU, CMU, and NCSU, although we still need to determine the method of distribution allowed to us). If you have any questions, feel free to send me mail at: warlord@mit.edu Enjoy! -derek -- Derek Atkins, SB '93 MIT EE, G MIT Media Laboratory Member, MIT Student Information Processing Board (SIPB) Home page: http://www.mit.edu:8001/people/warlord/home_page.html warlord@MIT.EDU PP-ASEL N1NWH PGP key available ------------------------------ From: jfc@athena.mit.edu (John F Carr) Crossposted-To: alt.filesystems.afs Subject: Re: Developing Distributed Filesystems for Linux? Date: 11 Sep 1994 19:56:31 GMT In article , Derrick J. Brashear wrote: >I too would be willing to work on a project like this, but as to whether >the answer is to write "free" AFS client software, or to write a totally >new system, I'm not sure. A reason I haven't put a lot of effort into my AFS work is because I don't like AFS. The only value of AFS is interoperation; if you want a distributed filesystem you can do better (write your own, or talk to some of the OS research groups). Part of the reason I started working on AFS was to get experience I could use to write something good later. An idea I've been thinking about for a while is to write a network filesystem that normally talks its own protocol but understands AFS or a useful subset of AFS. So ask, what is the target customer for your filesystem software? Are you trying to link Linux users together, or trying to make Linux work better in an AFS environment? If you are looking for a distributed filesystem without concern for compatibility, do you care about non-Linux systems? -- John Carr (jfc@mit.edu) ------------------------------ From: richard@atheist.tamu.edu (Richard Henderson) Subject: Re: Alpha Linux Date: 11 Sep 1994 18:55:42 GMT In article <34n3mv$k0n@news.tuwien.ac.at>, Anton Ertl wrote: >Why is this a good thing? And is it also a good thing that programs >break that assume that an index into an array or the difference >between to pointers fits into an int? That is exactly what the ANSI types size_t and ptrdiff_t are for -- use them. >And if you say, that such things can be easily fixed by greping for >"int" and replacing it with "TrueInt", you are wrong. In C int is not >just another type, it affects many things in the language: Normal >integer constants have the int type, int plays a special role in >conversion rules, switch wants an int argument, etc. No, switch wants a scalar. And what do you mean by "normal" integer constants? With sizeof(int) == 2, sizeof(long) == 4, 32768 unadorned with "L" is a long constant. >I have debugged enough programs (one:-) that worked perfectly well on >decent OSs on an Atari ST to know why I dislike >sizeof(int)!=sizeof(void *). Just writing "3" instead of "3L" cost at >least 5 hours of debugging time of two people (one of them an >experienced Atari user). I'll hazzard a guess that the problem of 3 vs 3L likely came from not prototyping properly? In short, there aren't nearly as many problems with sizeof(int) != sizeof(void*) as you'd like to believe. r~ ------------------------------ ** 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 ****************************** .