, so-and-so writes:". Also cut the original article down as much as possible; just keep enough of it to remind readers of the context. - Keep lines in posted articles to 72-75 characters. Many newsreaders chop off column 81 or arbitrarily insert a newline there, which makes longer lines difficult or impossible to read. But you need to keep well below 80 characters per line to allow for the > characters that get inserted when other people post follow-ups to your article. - Keep your signature to 4 lines or less (including any graphics), and for heaven's sake make sure it doesn't get posted twice in your article. - Don't post email without first obtaining the permission of the sender. Subject: 107. How can I learn more about Usenet? (new: 7 Aug 1993) There are two important newsgroups for learning about how Usenet and newsreader software works: - news.announce.newusers contains periodic postings that everybody is asked to read before posting anything to Usenet. (In theory, all new users are subscribed to news.announce.newusers automatically. But in practice not all newsreader software does that, so that many people violate the guidelines given there simply because they don't know about them.) - news.newusers.questions is described as "Q & A for users new to the Usenet". But new and long-time users can ask or answer questions about Usenet and newsreader software there. There's an important article, "Welcome to news.newusers.questions! (weekly posting)", that everyone is asked to read before posting to news.newusers.questions. (See below for ways to get a copy of that article.) The following postings in news.announce.newusers might be considered the "mandatory course" for new users: Introduction to news.announce.newusers What is Usenet? Answers to Frequently Asked Questions Rules for posting to Usenet A Primer on How to Work With the Usenet Community Hints on writing style for Usenet Emily Postnews Answers Your Questions on Netiquette The articles mentioned above are downloadable via ftp from rtfm.mit.edu in the following files: /pub/usenet/news.answers/news-newusers-intro Welcome to news.newusers.questions! (weekly posting) /pub/usenet/news.answers/news-announce-intro/part1 Introduction to news.announce.newusers /pub/usenet/news.answers/what-is-usenet/part1 What is Usenet? /pub/usenet/news.answers/usenet-faq/part1 Answers to Frequently Asked Questions /pub/usenet/news.answers/posting-rules/part1 Rules for posting to Usenet /pub/usenet/news.answers/usenet-primer/part1 A Primer on How to Work With the Usenet Community /pub/usenet/news.answers/usenet-writing-style/part1 Hints on writing style for Usenet /pub/usenet/news.answers/emily-postnews/part1 Emily Postnews Answers Your Questions on Netiquette They are also available from the mail server whose address is mail-server@rtfm.mit.edu. Send a message containing one or more "send" command lines such as send usenet/news.answers/news-newusers-intro The "send" filenames are the same as the ftp filenames except the "/pub" at the beginning is omitted. To get general information about the mail server, send it a message containing "help". Subject: 108. What other technical newsgroups should I know about? (rev: 16 Aug 1993) Caution: Some of these newsgroups have specialized charters; you'll probably get (and deserve) some flames if you post to an inappropriate group. Most groups have FAQ lists that will tell you what's appropriate. Don't post a request for the FAQ list; instead, retrieve it yourself--see "Where are FAQ lists archived?", in section C, "More information". - misc.forsale.computers.d and misc.forsale.computers.pc-clone are where you post notices of equipment, software, or computer books that you want to sell. Please don't post or crosspost those notices to comp.os.msdos.programmer. - comp.os.ms-windows.programmer.tools and ...misc (formerly part of comp.windows.ms.programmer): Similar to this group, but focus on programming for the MS-Windows platform. - comp.sys.ibm.pc.hardware is for more hardware-oriented discussions of the machines that run DOS. - the various comp.lang.* groups for articles and questions on the programming languages. Caution: some groups welcome discussions that are operating-system dependent or vendor specific; others do not. For example, comp.lang.c is definitely _not_ for questions about programming DOS or PC system features, even if the programs are written in C. On the other hand, comp.lang.pascal contains much discussion that is specific to Turbo Pascal. - comp.binaries.ibm.pc.wanted: AFTER you have looked in the other groups, this is the place to post a request for a particular binary program. - comp.archives.msdos.announce (moderated) explains how to use the archive sites, especially Garbo and Simtel, and lists files uploaded to them. Discussions belong in comp.archives.msdos.d, which replaced comp.binaries.ibm.pc.archives in December 1992. - MSDOS-Ann mailing list, for those who cannot subscribe to the comp.archives.msdos.announce newsgroup, lists Simtel and Garbo uploads in digest format. Instructions are downloadable as /pc/MSDOSANN.ZIP from Garbo pd1:msdosann.zip from Simtel. - comp.binaries.ibm.pc.d is for discussions about programs posted in comp.binaries.ibm.pc, and only those programs. This is a good place to report bugs in the programs, but not to ask where to find them (see cbip.wanted, above). cbip.d is NOT supposed to be a general PC discussion group. - comp.sources.misc: a moderated group for source code for many computer systems. It tends to get lots of Unix stuff, but you may also pick up some DOS-compatible code here. - alt.sources: an unmoderated group for source code. Guidelines are posted periodically. - Turbo Vision is a mailing list, not a newsgroup; send email to listserv@vtvm1.cc.vt.edu if you want to subscribe. section 2. Compile and link =========================== Subject: 201. What the heck is "DGROUP > 64K"? (rev: 18 Apr 1993) DGROUP is a link-time group of data segments, and the compiler typically generates code that expects DS to be pointing to DGROUP. (Exception: Borland's huge model has no DGROUP.) Here's what goes into DGROUP: - tiny model (all pointers near): DGROUP holds the entire program. - small and medium models (data pointers near): DGROUP holds all globals and static variables including string literals, plus the stack and the heap. - large, compact, and huge models in Microsoft (data pointers far): DGROUP holds only initialized globals and static variables including string literals, plus the stack and the near heap. - large and compact models in Borland (data pointers far): DGROUP holds initialized and uninitialized globals and static variables including string literals, but not the stack or heap. - huge model in Borland (data pointers far): there is no DGROUP, so the 64K limit doesn't apply. In all of the above, which is to say all six models in Microsoft C and all but huge in Borland C, DGROUP is limited to 64K including string literals (which are treated as static data). This limitation is due to the Intel CPU's segmented architecture. See the next Q for possible remedies. For more information, see topics like "memory models" and "memory management" in the index of your compiler manual. Also see ti738.asc, downloadable as part of pd1:bchelp10.zip at Simtel /pc/turbopas/bchelp10.zip at Garbo, for an extended general discussion of memory usage in Borland C programs, of which much applies to any C compiler in DOS. Subject: 202. How do I fix "automatic data segment exceeds 64K" or "stack plus data exceed 64K"? These messages are a variation of "DGROUP > 64K". For causes, please see the preceding Q. If you get this error in tiny model, your program is simply too big and you must use a different memory model. If you get this link error in models S, C, M, L, or Microsoft's H, there are some things you can do. (This error can't occur in Borland's huge model.) If you have one or two big global arrays, simply declare them far. The compiler takes this to mean that any references to them will use 32-bit pointers, so they'll be in separate segments and no longer part of DGROUP. Or you can use the /Gt[number] option with Microsoft or -Ff[=size] with Borland C++ 2.0 and up. This will automatically put variables above a certain size into their own segments outside of DGROUP. Yet another option is to change global arrays to far pointers. Then at the beginning of your program, allocate them from the far heap (_fmalloc in Microsoft, farmalloc in Borland). Finally, you can change to huge model (with Borland compilers, not Microsoft). Borland's H model still uses far pointers by default, but "sets aside the [64K] limit" and has no DGROUP group, according to the BC++ 2.0 Programmer's Guide. Microsoft's H model does use huge data pointers by default but retains DGROUP and its 64K limit, so switching to the H model doesn't buy you anything if you have DGROUP problems. Subject: 203. Will Borland C code and Microsoft C code link together? (rev: 15 Aug 1993) Typically this question is asked by someone who owns compiler A and is trying to write code to link with a third-party library that was compiled under compiler B. The answer to the question is, Not in general. Here are some of the reasons: - "Helper" functions (undocumented functions for stack checking, floating-point arithmetic, and operations on longs) differ between the two compilers. - The compilers may embed instructions in the object code that tell the linker to look for their own run-time libraries. You can use the linker option that says to ignore such instructions: /n in TLINK, /NOD in the Microsoft linker (the one that comes with the C compiler, not the one that used to come with DOS). But getting around this problem will very likely just reveal other problems, like different helper functions, that have no easy solution. Those problems will generate link-time errors. Others may not show up until run time: - Borland's compact, large, and huge models don't assume DS=SS, but Microsoft's do. The -Fs option on the Borland compiler, or one of the /A options on Microsoft, should take care of this problem -- once you know that's what's going on. - Check conventions for ordering and packing structure members, and for alignment of various types on byte, word, paragraph, or other boundaries. Again, you can generally adjust your code to match if you know what conventions were used in compiling the "foreign" libraries. - Check the obvious and make sure that your code was compiled under the same memory model as the code you're trying to link with. (That's necessary, but no guarantee. Microsoft and Borland don't use exactly the same conventions for segments and groups, particularly in the larger memory models.) That said, there are some circumstances where you can link hybrids. Your best chance of success comes if you compile in large model with the compiler switch that says to reload DS on entry to each function, avoid longs and floating point, use only 16-bit pointers, suppress stack checking, and specify all libraries used in the link. Subject: 204. Why did my program bomb at run time with "floating point formats not linked"? First, is that the actual message, or did it say "floating point not loaded"? If it was the latter, see the next Q. You're probably using a Borland compiler for C or C++ (including Turbo C and Turbo C++). Borland's compilers try to be smart and not link in the floating-point (f-p) library unless you need it. Alas, they all get the decision wrong. One common case is where you don't call any f-p functions, but you have %f or other f-p formats in scanf/printf calls. The cure is to call an f-p function, or at least force one to be present in the link. To do that, define this function somewhere in a source file but don't call it: static void forcefloat(float *p) { float f = *p; forcefloat(&f); } It doesn't have to be in the module with the main program, as long as it's in a module that will be included in the link. If you have Borland C++ 3.0, the README file documents a slightly less ugly work-around. Insert these statements in your program: extern unsigned _floatconvert; #pragma extref _floatconvert Subject: 205. Why did my program bomb with "floating point not loaded"? That is Microsoft C's run-time message when the code requires a numeric coprocessor but your computer doesn't have one installed. If the program is yours, relink it using the xLIBCE or xLIBCA library (where x is the memory model). Subject: 206. How can I change the stack size in Borland's C compilers? In Turbo C, Turbo C++, and Borland C++, you may not find "stack size" in the index but the global variable _stklen should be there. The manual will instruct you to put a statement like extern unsigned _stklen = 54321U; in your code, outside of any function. You must assign the value right in the extern statement; it won't work to assign a value at run time. (The "extern" in this context isn't ANSI C and ought not to be required, but the above statement is a direct quote from the Library Reference manual of Borland C++ 2.0.) The linker may give you a duplicate symbol warning, which you can ignore. Subject: 207. What's the format of an .OBJ file? (rev: 24 Sep 1993) - base .OBJ format: Intel's document number #121748-001, {8086 Relocatable Object Module Formats}. (not verified) Both Microsoft and Borland have extended the .OBJ format, as has IBM for OS/2; and according to the MS-DOS encyclopedia, Microsoft doesn't actually use all the listed formats. - Microsoft-specific .OBJ formats: * The .OBJ format document dated 14 Dec 1992, a WinHelp file from the Microsoft developer's CD-ROM (266K after unzipping), is downloadable as /vendor/microsoft/developer-network/ctech/11-9.zip from ftp.uu.net. (verified by SB) * A 45-page article can be found in the {MS-DOS Encyclopedia}, ISBN 1-55615-049-0 (verified by SB), now out of print. * "Microsoft Object Module Format (OMF)" Specification, 22 Nov 1991, was published by the Microsoft Languages Group. (not verified) - Borland-specific .OBJ formats: Open Architecture Handbook. The Borland Developer's Technical Guide, 1991, no ISBN. Chapter 2, "Object file contents", (pages 27-50) covers the comment records sent to the object file by Borland C++ version 3.0 and other Borland compilers. The comment records mostly contain information for the Borland debugger. (not verified) - A "tutorial on the .OBJ format" comes with the VAL experimental linker, downloadable as pd1:val-link.arc at Simtel /pc/assembler/linker.zoo at Garbo. Despite such different names, those files have the same contents, but their contents are dated 18 Feb 1989. You'd be better off with one of the more recent references listed above. Subject: 208. What's the format of an .EXE header? See pages 349-350 of {PC Magazine} 30 June 1992 (xi:12) for the old and new formats. For a more detailed layout, look under INT 21 function 4B in Ralf Brown's interrupt list. Ralf Brown's list includes extensions for Borland's TLINK and Borland debugger info. Among the books that detail formats of executable files are {DOS Programmer's Reference: 2d Edition} by Terry Dettman and Jim Kyle, ISBN 0-88022-458-4; and {Microsoft MS-DOS Programmer's Reference}, ISBN 1-55615-329-5. Subject: 209. What's the difference between .COM and .EXE formats? To oversimplify: a .COM file is a direct image of core, and an .EXE file will undergo some further relocation when it is run (and so it begins with a relocation header). A .COM file is limited to 64K for all segments combined, but an .EXE file can have as many segments as your linker will handle and be as large as RAM can take. The actual file extension doesn't matter. DOS knows that a file being loaded is in .EXE format if its first two bytes are MZ or ZM; otherwise it is assumed to be in .COM format. For instance, DR-DOS 6.0's COMMAND.COM is in .EXE format. section 3. Keyboard =================== Subject: 301. How can I read a character without echoing it to the screen, and without waiting for the user to press the Enter key? The C compilers from Microsoft and Borland offer getch (or getche to echo the character); Turbo Pascal has ReadKey. In other programming languages, load 8 in register AH and execute INT 21; AL is returned with the character from standard input (possibly redirected). If you don't want to allow redirection, or you want to capture Ctrl-C and other special keys, use INT 16 with AH=10; this will return the scan code in AH and ASCII code (if possible) in AL, except that AL=E0 with AH nonzero indicates one of the grey "extended" keys was pressed. (If your BIOS doesn't support the extended keyboard, use INT 16 function 0 not 10.) Subject: 302. How can I find out whether a character has been typed, without waiting for one? In Turbo Pascal, use KeyPressed. Both Microsoft C and Turbo C offer the kbhit( ) function. All of these tell you whether a key has been pressed. If no key has been pressed, they return that information to your program. If a keystroke is waiting, they tell your program that but leave the key in the input buffer. You can use the BIOS call, INT 16 function 01 or 11, to check whether an actual keystroke is waiting; or the DOS call, INT 21 function 0B, to check for a keystroke from stdin (subject to redirection). See Ralf Brown's interrupt list. Subject: 303. How can I disable Ctrl-C/Ctrl-Break and/or Ctrl-Alt-Del? (rev: 11 Sep 1993) Several utilities are downloadable from pd1: at Simtel. In that directory, cadel.zip contains a TSR (with source code) to disable those keys. Also, keykill.arc contains two utilities: keykill.com lets you disable up to three keys of your choice, and deboot.com changes the boot key to leftShift-Alt-Del. C programmers who simply want to make sure that the user can't Ctrl-Break out of their program can use the ANSI-standard signal( ) function; the Borland compilers also offer ctrlbrk( ) for handling Ctrl-Break. However, if your program uses normal DOS input such as getch( ), ^C will appear on the screen when the user presses Ctrl-C or Ctrl-Break. You can avoid the ^C echo for Ctrl-C by using _bios_keybrd( ) in MSC or bioskey( ) in BC++; however, Ctrl-Break will still terminate the program. An alternative approach involves programming input at a lower level. You can use INT 21 function 7, which allows redirection but doesn't echo the ^C (or any other character, for that matter); or use INT 16 function 0 or 10; or hook INT 9 to discard Ctrl-C and Ctrl-Break before the regular BIOS keyboard handler sees them; etc., etc. You should be aware that Ctrl-C and Ctrl-Break are processed quite differently internally. Ctrl-Break, like all keystrokes, is processed by the BIOS code at INT 9 as soon as the user presses the keys, even if earlier keys are still in the keyboard buffer: by default the handler at INT 1B is called. Ctrl-C is not special to the BIOS, nor is it special to DOS functions 6 and 7; it _is_ special to DOS functions 1 and 8 when at the head of the keyboard buffer. You will need to make sure BREAK is OFF to prevent DOS polling the keyboard for Ctrl-C during non-keyboard operations. Some good general references are {Advanced MS-DOS} by Ray Duncan, ISBN 1-55615-157-8; {8088 Assembler Language Programming: The IBM PC}, ISBN 0-672-22024-5, by Willen & Krantz; and {COMPUTE!'s Mapping the IBM PC}, ISBN 0-942386-92-2. Subject: 304. How can I disable the print screen function? There are really two print screen functions: 1) print current screen snapshot, triggered by PrintScreen or Shift-PrtSc or Shift-grey*, and 2) turn on continuous screen echo, started and stopped by Ctrl-P or Ctrl-PrtSc. 1) Screen snapshot to printer The BIOS uses INT 5 for this. Fortunately, you don't need to mess with that interrupt handler. The standard handler, in BIOSes dated December 1982 or later, uses a byte at 0040:0100 (alias 0000:0500) to determine whether a print screen is currently in progress. If it is, pressing PrintScreen again is ignored. So to disable the screen snapshot, all you have to do is write a 1 to that byte. When the user presses PrintScreen, the BIOS will think that a print screen is already in progress and will ignore the user's keypress. You can re-enable PrintScreen by zeroing the same byte. Here's some simple code: void prtsc_allow(int allow) /* 0=disable, nonzero=enable */ { unsigned char far* flag = (unsigned char far*)0x00400100UL; *flag = (unsigned char)!allow; } 2) Continuous echo of screen to printer If ANSI.SYS is loaded, you can easily disable the continuous echo of screen to printer (Ctrl-P or Ctrl-PrtSc). Just redefine the keys by "printing" strings like these to the screen (BASIC print, C printf, Pascal Write statements, or ECHO command in batch files): <27>[0;114;"Ctrl-PrtSc disabled"p <27>[16;"^P"p Change <27> in the above to an Escape character, ASCII 27. If you haven't installed ANSI.SYS, I (SB) can't offer an easy way to disable the echo-screen-to-printer function. Please send any tested solutions to the editor. Actually, you might not need to disable Ctrl-P and Ctrl-PrtSc. If your only concern is not locking up your machine, when you see the "Abort, Retry, Ignore, Fail" prompt just press Ctrl-P again and then I. As an alternative, install one of the many print spoolers that intercept printer-status queries and always return "Printer ready". Subject: 305. How can my program turn NumLock (CapsLock, ScrollLock) on or off? First, if you just don't want NumLock turned on when you reboot, check your system's setups. (Use Ctrl-Alt-Enter any time, or press a special key like Del at boot time, or run the setup program supplied with your system.) Many systems now have an option in setup to turn NumLock off at boot time. You need to twiddle bit 5, 6, or 4 of location 0040:0017. Here's some code: lck( ) turns on a lock state, and unlck( ) turns it off. (The status lights on some keyboards may not reflect the change. If yours is one, call INT 16 function 2, "get shift status", and that may update them. It will certainly do no harm.) #define NUM_LOCK (1 << 5) #define CAPS_LOCK (1 << 6) #define SCRL_LOCK (1 << 4) void lck(int shiftype) { char far* kbdstatus = (char far*)0x00400017UL; *kbdstatus |= (char)shiftype; } void unlck(int shiftype) { char far* kbdstatus = (char far*)0x00400017UL; *kbdstatus &= ~(char)shiftype; } Subject: 306. How can I speed up the keyboard's auto-repeat? The keyboard speed has two components: delay (before a key that you hold down starts repeating) and typematic rate (the speed once the key starts repeating). Most BIOSes since 1986 let software change the delay and typematic rate by calling INT 16 function 3, "set typematic rate and delay"; see Ralf Brown's interrupt list. If you have DOS 4.0 or later, you can use the MODE CON command that you'll find in your DOS manual. On 83-key keyboards (mostly XTs), the delay and typematic rate can't easily be changed. According to the {PC Magazine} of 15 Jan 1991, page 409, to adjust the typematic rate you need "a memory-resident program which simply '[watches]' the keyboard to see if you're holding down a key ... and after a certain time [starts] stuffing extra copies of the held-down key into the buffer." No source code is given in that issue; but the QUICKEYS utility that {PC} published in 1986 does this sort of watching (not verified); source and object code are downloadable in pd1:vol5n05.arc from Simtel. Subject: 307. What is the SysRq key for? There is no standard use for the key. The BIOS keyboard routines in INT 16 simply ignore it; therefore so do the DOS input routines in INT 21 as well as the keyboard routines in libraries supplied with high-level languages. When you press or release a key, the keyboard triggers hardware line IRQ1, and the CPU calls INT 9. INT 9 reads the scan code from the keyboard and the shift states from the BIOS data area. What happens next depends on whether your PC's BIOS supports an enhanced keyboard (101 or 102 keys). If so, INT 9 calls INT 15 function 4F to translate the scan code. If the translated scan code is 54 hex (for the SysRq key) then INT 9 calls INT 15 function 85 and doesn't put the keystroke into the keyboard buffer. The default handler of that function does nothing and simply returns. (If your PC has an older BIOS that doesn't support the extended keyboards, INT 15 function 4F is not called. Early ATs have 84-key keyboards, so their BIOS calls INT 15 function 85 but nor 4F.) Thus your program is free to use SysRq for its own purposes, but at the cost of some programming. You could hook INT 9, but it's probably easier to hook INT 15 function 85, which is called when SysRq is pressed or released. Subject: 308. How can my program tell what kind of keyboard is on the system? Ralf Brown's Interrupt List includes MEMORY.LST, a detailed breakdown by Robin Walker of the contents of the BIOS system block that starts at 0040:0000. Bit 4 of byte 0040:0096 is "1=enhanced keyboard installed". C code to test the keyboard type: char far *kbd_stat_byte3 = (char far *)0x00400096UL; if (0x10 & *kbd_stat_byte3) /* 101- or 102-key keyboard is installed */ {PC Magazine} 15 Jan 1991 issue suggests on page 412 that "for some clones [the above test] is not foolproof". If you use this method in your program you should provide the user some way to override this test, or at least some way to tell your program to assume a non-enhanced keyboard. The article suggests a different approach to determining the type of keyboard. Subject: 309. How can I tell if input, output, or stderr has been redirected? (rev: 24 Sep 1993) Normally, input and output are associated with the console (i.e., with the keyboard and the screen, respectively). If either is not, you know that it has been redirected. Some source code to check this is available at the usual archive sites. If you program in Turbo Pascal, you'll want this downloadable collection of Turbo Pascal units: /pc/ts/tspa33*.zip at Garbo pd1:tspa33*.zip at Simtel. (where the * is 70, 60, 55, 50, or 40 for Turbo Pascal 7.0, 6.0, 5.5, 5.0, or 4.0 respectively.) Source code is not included. Also see the downloadable Frequently Asked Questions files by Timo Salmi: /pc/ts/tsfaqp15.zip at Garbo pd1:tsfaqp15.zip at Simtel. If you program in C, use isatty( ) if your implementation has it. Otherwise, pd1:is_con10.zip is downloadable from Simtel; it includes source code. Good references for the principles are {PC Magazine} 16 Apr 1991 (x:7) pg 374; Ray Duncan's {Advanced MS-DOS}, ISBN 1-55615-157-8, or Ralf Brown's interrupt list for INT 21 function 4400; and Terry Dettman and Jim Kyle's {DOS Programmer's Reference: 2d edition}, ISBN 0-88022-458-4, pp 602-603. Subject: 310. How can I increase the size of the keyboard buffer? (new: 20 June 1993) SB tested only one of the many available device drivers that do this, namely BUF160, which extends the keyboard buffer to 160 characters. It performed flawlessly for two years with MS-DOS 5 and Windows 3.1. It's downloadable as pd1:buf160_6.zip at Simtel /pc/keyboard/buf160_6.zip at Garbo. Subject: 311. How can I stuff characters into the keyboard buffer? (new: 15 Aug 1993) If your computer has an enhanced keyboard (see "How can my program tell what kind of keyboard is on the system?", above), then put the scan code in CH, the ASCII character in CL, 5 in AH, and execute INT 16. The return in AL is 0 for success or 1 for buffer full. (continued in part 3) -- Stan Brown, Oak Road Systems brown@Ncoast.ORG Can't find FAQ lists? ftp to 'rtfm.mit.edu' and look in /pub/usenet (or email me >>> with valid reply-to address <<< for instructions).