otto@tukki.jyu.fi (Otto J. Makela) (01/03/91)
In a fit of desperation, I wrote the following code to get rid of trailing
SUB's (^Z's, 0x1A) from MeSsy-DOS files, to make porting programs from Unix
easier... what am I doing wrong ? Or is it the compiler (I tried compiling
this with an ancient Turbo C 2.0) ? I've made similar code to work on Manx
Aztec C86 4.10d (a nice compiler, but sadly out of date).
--
#include <stdio.h>
#include <fcntl.h>
/*
** MeSsy-DOS kludge for fopen-append
*/
FILE *fopena(filename)
char *filename; {
FILE *f;
if(f=fopen(filename,"a+b")) {
fseek(f,-1L,SEEK_END);
while(fgetc(f)==0x1A) fseek(f,-2L,SEEK_CUR);
/* This is needed to clear stdio buffering */
fseek(f,ftell(f),SEEK_SET);
/* Then set text mode on the stream */
setmode(fileno(f),O_TEXT);
}
return f;
}
main(argc,argv)
int argc;
char *argv[]; {
FILE *f;
if(argc<2)
fprintf(stderr,"usage: %s file_to_append_to\n",*argv);
else if(f=fopena(argv[1])) {
fprintf(f,"This is the appended text.\n");
fclose(f);
} else
fprintf(stderr,"%s: could not open %s for append\n",
*argv,argv[1]);
}
--
/* * * Otto J. Makela <otto@jyu.fi> * * * * * * * * * * * * * * * * * * */
/* Phone: +358 41 613 847, BBS: +358 41 211 562 (CCITT, Bell 24/12/300) */
/* Mail: Kauppakatu 1 B 18, SF-40100 Jyvaskyla, Finland, EUROPE */
/* * * Computers Rule 01001111 01001011 * * * * * * * * * * * * * * * * */wirzeniu@cs.Helsinki.FI (Lars Wirzenius) (01/04/91)
In article <OTTO.91Jan3125711@tukki.jyu.fi> otto@tukki.jyu.fi (Otto J. Makela) writes: > if(f=fopen(filename,"a+b")) { "a" means that all writes happen at the end of the file, use "r+b" instead. -- Lars Wirzenius wirzeniu@cs.helsinki.fi wirzenius@cc.helsinki.fi
otto@tukki.jyu.fi (Otto J. Makela) (01/04/91)
In article <10251@hydra.Helsinki.FI> wirzeniu@cs.Helsinki.FI (Lars Wirzenius) writes: In article <OTTO.91Jan3125711@tukki.jyu.fi> otto@tukki.jyu.fi (Otto J. Makela) writes: > if(f=fopen(filename,"a+b")) { "a" means that all writes happen at the end of the file, use "r+b" instead. ALL WRITES ? Come on, the first thing I do after that is to seek the file... I'd like to point out I can't use "r+b", 'cause it'll fail if the file does not exist beforehand. -- /* * * Otto J. Makela <otto@jyu.fi> * * * * * * * * * * * * * * * * * * */ /* Phone: +358 41 613 847, BBS: +358 41 211 562 (CCITT, Bell 24/12/300) */ /* Mail: Kauppakatu 1 B 18, SF-40100 Jyvaskyla, Finland, EUROPE */ /* * * Computers Rule 01001111 01001011 * * * * * * * * * * * * * * * * */
karl@ima.isc.com (Karl Heuer) (01/05/91)
In <OTTO.91Jan4115449@tukki.jyu.fi> otto@tukki.jyu.fi (Otto J. Makela) writes: >In <10251@hydra.Helsinki.FI> wirzeniu@cs.Helsinki.FI (Lars Wirzenius) writes: >>"a" means that all writes happen at the end of the file, use "r+b" instead. > >ALL WRITES ? Come on, the first thing I do after that is to seek the file... Lars is correct. On a file opened in append-mode, fseek() is only useful for reading. (Though some historical implementations implemented append-mode with an initial seek-to-end only, this is not acceptable in ANSI C.) >I'd like to point out I can't use "r+b", 'cause it'll fail if the file does >not exist beforehand. Then use "w+b". Karl W. Z. Heuer (karl@ima.isc.com or uunet!ima!karl), The Walking Lint
otto@tukki.jyu.fi (Otto J. Makela) (01/07/91)
In article <1991Jan04.173903.20643@dirtydog.ima.isc.com> karl@ima.isc.com (Karl Heuer) writes: [...] Lars is correct. On a file opened in append-mode, fseek() is only useful for reading. (Though some historical implementations implemented append- mode with an initial seek-to-end only, this is not acceptable in ANSI C.) Yes. Confirm. Affirmative. Now that I've gone thru it, this seems to be how it works. I didn't realize ANSI decided to break compatibility with Unix in such a radical way... >I'd like to point out I can't use "r+b", 'cause it'll fail if the file does >not exist beforehand. Then use "w+b". Can't do that either, 'cause I don't want to kill the file. I need both. Here is the fixed code, if someone finds it useful... -- #include <stdio.h> #include <fcntl.h> #ifdef UNIX #define fopena(filename) fopen(filename,"a") #else /* ** MeSsy-DOS kludge to replace fopen(filename,"at") */ FILE *fopena(filename) char *filename; { FILE *f; if(f=fopen(filename,"r+b")) { fseek(f,-1L,SEEK_END); while(fgetc(f)==0x1A) fseek(f,-2L,SEEK_CUR); /* This is needed to clear stdio buffering */ fseek(f,ftell(f),SEEK_SET); setmode(fileno(f),O_TEXT); } else f=fopen(filename,"wt"); return f; } #endif /* ** Test main program */ main(argc,argv) int argc; char *argv[]; { FILE *f; if(argc<2) fprintf(stderr,"usage: %s file_to_append_to\n",*argv); else if(f=fopena(argv[1])) { fprintf(f,"This is the appended text.\n"); fclose(f); } else fprintf(stderr,"%s: could not open %s for append\n", *argv,argv[1]); } -- /* * * Otto J. Makela <otto@jyu.fi> * * * * * * * * * * * * * * * * * * */ /* Phone: +358 41 613 847, BBS: +358 41 211 562 (CCITT, Bell 24/12/300) */ /* Mail: Kauppakatu 1 B 18, SF-40100 Jyvaskyla, Finland, EUROPE */ /* * * Computers Rule 01001111 01001011 * * * * * * * * * * * * * * * * */
gwyn@smoke.brl.mil (Doug Gwyn) (01/09/91)
In article <OTTO.91Jan7070856@tukki.jyu.fi> otto@tukki.jyu.fi (Otto J. Makela) writes: >In article <1991Jan04.173903.20643@dirtydog.ima.isc.com> karl@ima.isc.com (Karl Heuer) writes: > Lars is correct. On a file opened in append-mode, fseek() is only useful > for reading. (Though some historical implementations implemented append- > mode with an initial seek-to-end only, this is not acceptable in ANSI C.) >Yes. Confirm. Affirmative. Now that I've gone thru it, this seems to be >how it works. I didn't realize ANSI decided to break compatibility with Unix >in such a radical way... I get mighty tired of claims that "ANSI C breaks compatibility with ...". In fact there are instances of UNIX that use O_APPEND for stdio "a" modes. On some primitive UNIX implementations lacking O_APPEND, an imperfect simulation of appending was implemented by performing just the initial seek to EOF. A proper implementation on such systems would have had to seek to EOF before each write operation. Note that the notion of appending to a file does not work well with multiple concurrent (buffered) writers.