Conference noted::ultrix

5462.0. "fseek issues with Ultrix V4.4" by HYDRA::TALATINIAN () Mon Mar 17 1997 16:41

  I am working with a partner who has reported a problem trying to fseek to an
odd byte in a file causing the product to crash.  The following example
illustrates the problem.  This does work correctly on Digital UNIX.  Are there
any known bugs or patches available?  Thanks in advance.



/* D Miller 3/5/97 - Changes from 'original' version (i.e. repro attached
	to 88698):
		1. Had to null-terminate 'val' in read_chars after reading,
			else this wouldn't even work on Solaris
		2. Added a few simple read tests before testing Write-then-read. 
		3. Added success message in read code. */

#include <stdio.h>

FILE *fp;

void read_chars();
void write_chars();

main ( argv, argc)
int argc;
char *argv;
{
	int loc;

	/* Write a file containing 'A' (at byte 0) to 'M' (at byte 12),
		then change some letters to numbers */

	fp = fopen("/tmp/proof_data", "w+");

	/* write values A-Z */
	write_chars( 0, "ABCDEFGHIJKLM");

	/* Read E */
	loc = 'E'-'A';
	read_chars( loc, "E");

	/* Read F */
	loc = 'F'-'A';
	read_chars( loc, "F");

	/* A->1 */
	loc = 0;
	read_chars( loc, "A");
	write_chars( loc, "1");
	read_chars( loc, "1");

	/* Add N at end */
	loc = 'N'-'A';
	write_chars( loc, "N");
	read_chars( loc, "N");

	/* B->2 */
	loc = 'B'-'A';
	read_chars( loc, "B");
	write_chars( loc, "2");
	read_chars( loc, "2");


	fprintf(stdout, "*** End of test - v4  ***\n");
	fclose(fp);
}

void read_chars( at, expect)
int			at;				/* Where to read */
char* expect;			/* Value expected there */
{
	int loc;
	int nitems;
	int size;
	char *val;

	/* Note we do a seek before an fread! */
	fseek( fp, (long)at, 0);

	loc = ftell(fp);
	if (loc != at) {
		fprintf(stdout, "ERROR in read_chars: Seek to %ld, but at %ld\n",
						at, loc);
	}

	size = strlen(expect);
	val = (char*)malloc(size+1);
	nitems = fread((char*)(val),size,1,fp);
	val[size] = '\0';
	if (nitems != 1) {
		fprintf(stdout, "ERROR in read_chars: Expected 1 item, but read
%d\n",
						nitems);
	}
	if (strcmp(val, expect)) {
		fprintf(stdout, "ERROR in read_chars: Expected '%s', but saw
'%s'\n",
						expect, val);
	}
	else {
		fprintf(stdout, "--- Successfully read '%s' at %ld\n",
						val, at);
	}
}


void write_chars( at, new)
int			at;				/* Where to write */
char *new;		/* Value to write */
{
	int loc;
	unsigned int val;
	int nitems;
	int size;

	/* Note we do a seek before a fwrite! */
	fseek( fp, (long)at, 0);

	loc = ftell(fp);
	if (loc != at) {
		fprintf(stdout, "ERROR in write_chars: Seek to %ld, but at
%ld\n",
						at, loc);
	}

	size = strlen(new);
	fprintf(stdout, "--- about to write %d chars at %ld\n", size,loc);
	fwrite((char*)new, size, 1, fp);

	/* We should be size bytes further in the file */
	loc = ftell(fp);
	if (loc != (at+size)) {
		fprintf(stdout, "ERROR after write_chars: Expected to be at %ld,
but at %ld\n",
						at+size, loc);
	}
}

	I don't know where the ULTRIX patch directory has gotten
	to.  It wasn't there I thought it was (guru::).  If you can
	find the V4.5 Release Notes it will have a list of all the
	problems that it fixed.  That should be on oskits:: some-
	where.

	Failing either of those, give the CSC a call and failing
	that start an IPMT case.

>	I don't know where the ULTRIX patch directory has gotten
>	to.  It wasn't there I thought it was (guru::).  

The patch area is guru::/usr/guest/patches/risc/v4.4 and
		  guru::/usr/guest/patches/vax/v4.4.

You should take the README file and see if the problem description there
matches what you are seeing.  If not, then it would seem an IPMT case is
needed.

-Joe Melvin
 USEG/ULTRIX