Trailing-Edge
-
PDP-10 Archives
-
SRI_NIC_PERM_FS_1_19910112
-
c/user/make/vaxvms.c
There are no other files named vaxvms.c in the archive.
/***********************************************************************
This file provides alternative functions for several VMS VMS C library
routines which either unacceptable, or incorrect, implementations. They
have been developed and tested under VMS Version 4.4, but indications
are that they apply to earlier versions, back to 3.2 at least. They
should be retested with each new release of VMS C.
Contents:
EXIT
FSEEK
FTELL
GETENV
READ
UNGETC
getlogin
qsort
system
tell
unlink
The VAX VMS file system record structure has unfortunate consequences
for random access files.
By default, text files written by most system utilities, and languages
other than C, have a variable length record format, in which a 16-bit
character count is aligned on an even-byte boundary in the disk block
b(always 512 bytes in VMS, independent of record and file formats),
followed by <count> bytes of data. Binary files, such as .EXE, .OBJ,
and TeX .DVI and font files, all use a 512-byte fixed record format
which has no explicit length field. No file byte count is stored;
instead, the block count, and the offset of the last data byte in the
last block are recorded in the file header (do ``DUMP/HEADER filespec''
to see it). For binary files with fixed-length records, the last block
is normally assumed to be full, and consequently, file transfer of
binary data from other machines via Kermit, FTP, or DCL COPY from ANSI
tapes, generally fails because the input file length is not a multiple
of 512.
This record organization may be contrasted with the STREAM, STREAM_LF,
and STREAM_CR organizations supported from Version 4.0; in these, disk
blocks contain a continuous byte stream in which nothing, or LF, or CR,
is recognized as a record terminator. These formats are similar to the
Unix and TOPS-20 file system formats which also use continuous byte
streams.
For C, this means that a program operating on a file in record format
cannot count input characters and expect that count to be the same value
as the offset parameter passed to fseek(), which numerous C programs
assume to be the case. The draft ANSI C standard, and Harbison and
Steele's ``C Reference Manual'', emphasize that only values returned by
ftell() should be used as arguments to fseek(), allowing the program to
return to a position previously read or written. UNFORTUNATELY, VMS C
ftell() DOES NOT RETURN A CORRECT OFFSET VALUE FOR RECORD FILES.
Instead, for record files, it returns the byte offset of the start of
the current record, no matter where in that record the current position
may be. This misbehavior is completely unnecessary, since the
replacements below perform correctly, and are written entirely in C.
Another problem is that ungetc(char c, FILE* fp) is unreliable. VMS C
implements characters as signed 8-bit integers (so do many other C
implementations). fgetc(FILE* fp) returns an int, not a char, whose
value is EOF (-1) in the event of end-of-file; however, this value will
also be returned for a character 0xFF, so it is essential to use
feof(FILE *fp) to test for a true end-of-file condition when EOF is
returned. ungetc() checks the sign of its argument c, and if it is
negative (which it will be for 128 of the 256 signed bytes), REFUSES TO
PUT IT BACK IN THE INPUT STREAM, on the assumption that c is really EOF.
This too can be fixed; ungetc() should only do nothing if feof()
indicates a true end-of-file condition. The overhead of this is
trivial, since feof() is actually implemented as a macro which does
nothing more than a logical AND and compare-with-zero.
Undoubtedly other deficiencies in VMS C will reveal themselves.
VMS read() returns only a single disk block on each call. Its
replacment, vms_read(), will return the requested number of bytes, if
possible.
There are also a few Unix standard functions which are unimplemented.
qsort() is not provided. getlogin() and unlink() have VMS equivalents
provided below. tell() is considered obsolete, since its functionality
is available from lseek(), but it is still seen in a few programs, so is
provided below. getenv() fails if the name contains a colon; its
replacement allows the colon.
In the interest of minimal source perturbation, replacements for VMS
functions are given the same names, but prefixed "vms_". For
readability, the original names are preserved, but are converted to
upper-case:
#define FTELL vms_ftell
#define FSEEK vms_fseek
#define GETENV vms_getenv
#define UNGETC vms_ungetc
These are only defined to work correctly for fixed length 512-byte
records, and no check is made that the file has that organization (it is
possible, but not without expensive calls to fstat(), or access to
internal library structures).
[17-Dec-86] -- Nelson H.F. Beebe, University of Utah Center for
Scientific Computing
***********************************************************************/
#define EXIT vms_exit
#define FTELL vms_ftell
#define FSEEK vms_fseek
#define GETENV vms_getenv
#define READ vms_read
#define UNGETC vms_ungetc
#include <stdio.h>
#include <types.h>
#include <ctype.h>
#include <stat.h>
#include <descrip.h>
#include <ssdef.h>
/**********************************************************************/
/*-->EXIT*/
void
vms_exit(code)
int code;
{
switch (code)
{
case 0:
exit(1); /* success */
break;
default:
exit(2); /* error */
break;
}
}
/**********************************************************************/
/*-->FSEEK*/
/* VMS fseek() and ftell() on fixed-length record files work correctly
only at block boundaries. This replacement code patches in the offset
within the block. Directions from current position and from
end-of-file are converted to absolute positions, and then the code for
that case is invoked. */
long
FSEEK(fp,n,dir)
FILE *fp;
long n;
long dir;
{
long k,m,pos,val,oldpos;
struct stat buffer;
for (;;) /* loops only once or twice */
{
switch (dir)
{
case 0: /* from BOF */
oldpos = FTELL(fp); /* get current byte offset in file */
k = n & 511; /* offset in 512-byte block */
m = n >> 9; /* relative block number in file */
if (((*fp)->_cnt) && ((oldpos >> 9) == m)) /* still in same block */
{
val = 0; /* success */
(*fp)->_ptr = ((*fp)->_base) + k; /* reset pointers to requested byte */
(*fp)->_cnt = 512 - k;
}
else
{
val = fseek(fp,m << 9,0); /* move to start of requested 512-byte block */
if (val == 0) /* success */
{
(*fp)->_cnt = 0; /* indicate empty buffer */
(void)fgetc(fp); /* force refill of buffer */
(*fp)->_ptr = ((*fp)->_base) + k; /* reset pointers to requested byte */
(*fp)->_cnt = 512 - k;
}
}
return(val);
case 1: /* from current pos */
pos = FTELL(fp);
if (pos == EOF) /* then error */
return (EOF);
n += pos;
dir = 0;
break; /* go do case 0 */
case 2: /* from EOF */
val = fstat(fileno(fp),&buffer);
if (val == EOF) /* then error */
return (EOF);
n += buffer.st_size - 1; /* convert filesize to offset and */
/* add to requested offset */
dir = 0;
break; /* go do case 0 */
default: /* illegal direction parameter */
return (EOF);
}
}
}
/**********************************************************************/
/*-->FTELL*/
/* With fixed-length record files, ftell() returns the offset of the
start of block. To get the true position, this must be biased by
the offset within the block. */
long
FTELL(fp)
FILE *fp;
{
char c;
long pos;
long val;
if ((*fp)->_cnt == 0) /* buffer empty--force refill */
{
c = fgetc(fp);
val = UNGETC(c,fp);
if (val != c)
return (EOF); /* should never happen */
}
pos = ftell(fp); /* this returns multiple of 512 (start of block) */
if (pos >= 0) /* then success--patch in offset in block */
pos += ((*fp)->_ptr) - ((*fp)->_base);
return (pos);
}
/**********************************************************************/
/*-->READ*/
int
READ(file_desc,buffer,nbytes)
register int file_desc;
register char *buffer;
register int nbytes;
{
register int ngot;
register int left;
for ((left = nbytes, ngot = 0); left > 0; /* NOOP */)
{
ngot = read(file_desc,buffer,left);
if (ngot < 0)
return (-1); /* error occurred */
buffer += ngot;
left -= ngot;
}
return(nbytes-left);
}
/**********************************************************************/
/*-->UNGETC*/
long
UNGETC(c,fp) /* VMS ungetc() is a no-op if c < 0 (which is half the time!) */
char c;
FILE *fp;
{
if ((c == EOF) && feof(fp))
return (EOF); /* do nothing at true end-of-file */
else if ((*fp)->_cnt >= 512)/* buffer full--no fgetc() done in this block!*/
return (EOF); /* must be user error if this happens */
else /* put the character back in the buffer */
{
(*fp)->_cnt++; /* increase count of characters left */
(*fp)->_ptr--; /* backup pointer to next available char */
*((*fp)->_ptr) = c; /* save the character */
return (c); /* and return it */
}
}
/**********************************************************************/
/*-->getenv*/
char*
GETENV(name)
char* name;
{
char* p;
char* result;
char ucname[256];
p = ucname;
while (*name) /* VMS logical names must be upper-case */
{
*p++ = islower(*name) ? toupper(*name) : *name;
++name;
}
*p = '\0';
p = strchr(ucname,':'); /* colon in name? */
if (p == (char *)NULL) /* no colon in name */
result = getenv(ucname);
else /* try with and without colon */
{
result = getenv(ucname);
if (result == (char *)NULL)
{
*p = '\0';
result = getenv(ucname);
*p = ':';
}
}
return (result);
}
/**********************************************************************/
/*-->getlogin*/
char*
getlogin()
{
return ((char *)getenv("USER")); /* use equivalent VMS routine */
}
/**********************************************************************/
/*-->qsort*/
/***********************************************************************
TeXindex uses the standard Unix library function qsort() for
record sorting. Unfortunately, qsort() is not a stable sorting
algorithm, so input order is not necessarily preserved for equal
sort keys. This is important, because the sorting is
case-independent, while the actual entries may not be. For
example, the input
\entry{i}{22}{{\CODE{i}}}
\entry{i}{42}{{\CODE{i}}}
\entry{I}{41}{{\CODE{I}}}
\entry{I}{42}{{\CODE{I}}}
produces
\initial {I}
\entry {{\CODE{i}}}{22}
\entry {{\CODE{I}}}{41--42}
\entry {{\CODE{i}}}{42}
instead of the correct
\initial {I}
\entry {{\CODE{i}}}{22, 42}
\entry {{\CODE{I}}}{41--42}
We therefore provide this stable shellsort replacement for
qsort() based on the code given on p. 116 of Kernighan and
Ritchie, ``The C Programming Language'', Prentice-Hall (1978).
This has order N**1.5 average performance, which is usually
slower than qsort(). In the interests of simplicity, we make no
attempt to handle short sequences by alternative methods.
[07-Nov-86]
***********************************************************************/
#define BASE(i) &base[(i)*width]
void
qsort(base, nel, width, compar)
char base[]; /* start of data in memory */
int nel; /* number of elements to be sorted */
int width; /* size (in bytes) of each element */
int (*compar)(); /* comparison function */
{
int gap;
int i;
int j;
register int k; /* inner exchange loop parameters */
register char* p;
register char* q;
register char c;
for (gap = nel/2; gap > 0; gap /= 2)
{
for (i = gap; i < nel; i++)
{
for (j = i-gap; j >= 0; j -= gap)
{
p = BASE(j);
q = BASE(j+gap);
if ((*compar)(p,q) <= 0)
break; /* exit j loop */
else
{
for (k = 0; k < width; (++p, ++q, ++k))
{
c = *q;
*q = *p;
*p = c;
}
}
}
}
}
}
/**********************************************************************/
/*-->system*/
int
system(s)
char *s;
{
struct dsc$descriptor t;
t.dsc$w_length = strlen(s);
t.dsc$a_pointer = s;
t.dsc$b_class = DSC$K_CLASS_S;
t.dsc$b_dtype = DSC$K_DTYPE_T;
return (LIB$SPAWN(&t) == SS$_NORMAL) ? 0 : 127;
}
/**********************************************************************/
/*-->tell*/
long
tell(handle)
int handle;
{
return (lseek(handle,0L,1));
}
/**********************************************************************/
/*-->unlink*/
int
unlink(filename)
char *filename;
{
return (delete(filename)); /* use equivalent VMS routine */
}