2 * Copyright (C) 1984-2022 Mark Nudelman
4 * You may distribute under the terms of either the GNU General Public
5 * License or the Less License, as specified in the README file.
7 * For more information, see the README file.
12 * Low level character input from the input file.
13 * We use these special purpose routines which optimize moving
14 * both forward and backward from the current read pointer.
18 #if MSDOS_COMPILER==WIN32C
25 extern dev_t curr_dev
;
26 extern ino_t curr_ino
;
30 #include <sys/statfs.h>
33 typedef POSITION BLOCKNUM
;
35 public int ignore_eoi
;
38 * Pool of buffers holding the most recently used blocks of the input file.
39 * The buffer pool is kept as a doubly-linked circular list,
40 * in order from most- to least-recently used.
41 * The circular list is anchored by the file state "thisfile".
44 struct bufnode
*next
, *prev
;
45 struct bufnode
*hnext
, *hprev
;
52 unsigned int datasize
;
53 unsigned char data
[LBUFSIZE
];
55 #define bufnode_buf(bn) ((struct buf *) bn)
58 * The file state is maintained in a filestate structure.
59 * A pointer to the filestate is kept in the ifile structure.
61 #define BUFHASH_SIZE 1024
63 struct bufnode buflist
;
64 struct bufnode hashtbl
[BUFHASH_SIZE
];
74 #define ch_bufhead thisfile->buflist.next
75 #define ch_buftail thisfile->buflist.prev
76 #define ch_nbufs thisfile->nbufs
77 #define ch_block thisfile->block
78 #define ch_offset thisfile->offset
79 #define ch_fpos thisfile->fpos
80 #define ch_fsize thisfile->fsize
81 #define ch_flags thisfile->flags
82 #define ch_file thisfile->file
84 #define END_OF_CHAIN (&thisfile->buflist)
85 #define END_OF_HCHAIN(h) (&thisfile->hashtbl[h])
86 #define BUFHASH(blk) ((blk) & (BUFHASH_SIZE-1))
89 * Macros to manipulate the list of buffers in thisfile->buflist.
91 #define FOR_BUFS(bn) \
92 for (bn = ch_bufhead; bn != END_OF_CHAIN; bn = bn->next)
95 (bn)->next->prev = (bn)->prev; \
96 (bn)->prev->next = (bn)->next;
98 #define BUF_INS_HEAD(bn) \
99 (bn)->next = ch_bufhead; \
100 (bn)->prev = END_OF_CHAIN; \
101 ch_bufhead->prev = (bn); \
104 #define BUF_INS_TAIL(bn) \
105 (bn)->next = END_OF_CHAIN; \
106 (bn)->prev = ch_buftail; \
107 ch_buftail->next = (bn); \
111 * Macros to manipulate the list of buffers in thisfile->hashtbl[n].
113 #define FOR_BUFS_IN_CHAIN(h,bn) \
114 for (bn = thisfile->hashtbl[h].hnext; \
115 bn != END_OF_HCHAIN(h); bn = bn->hnext)
117 #define BUF_HASH_RM(bn) \
118 (bn)->hnext->hprev = (bn)->hprev; \
119 (bn)->hprev->hnext = (bn)->hnext;
121 #define BUF_HASH_INS(bn,h) \
122 (bn)->hnext = thisfile->hashtbl[h].hnext; \
123 (bn)->hprev = END_OF_HCHAIN(h); \
124 thisfile->hashtbl[h].hnext->hprev = (bn); \
125 thisfile->hashtbl[h].hnext = (bn);
127 static struct filestate
*thisfile
;
128 static int ch_ungotchar
= -1;
129 static int maxbufs
= -1;
134 extern int screen_trashed
;
135 extern int follow_mode
;
136 extern constant
char helpdata
[];
137 extern constant
int size_helpdata
;
138 extern IFILE curr_ifile
;
141 extern char *namelogfile
;
144 static int ch_addbuf();
148 * Get the character pointed to by the read pointer.
161 if (thisfile
== NULL
)
165 * Quick check for the common case where
166 * the desired char is in the head buffer.
168 if (ch_bufhead
!= END_OF_CHAIN
)
170 bp
= bufnode_buf(ch_bufhead
);
171 if (ch_block
== bp
->block
&& ch_offset
< bp
->datasize
)
172 return bp
->data
[ch_offset
];
178 * Look for a buffer holding the desired block.
180 h
= BUFHASH(ch_block
);
181 FOR_BUFS_IN_CHAIN(h
, bn
)
183 bp
= bufnode_buf(bn
);
184 if (bp
->block
== ch_block
)
186 if (ch_offset
>= bp
->datasize
)
188 * Need more data in this buffer.
194 if (bn
== END_OF_HCHAIN(h
))
197 * Block is not in a buffer.
198 * Take the least recently used buffer
199 * and read the desired block into it.
200 * If the LRU buffer has data in it,
201 * then maybe allocate a new buffer.
203 if (ch_buftail
== END_OF_CHAIN
||
204 bufnode_buf(ch_buftail
)->block
!= -1)
207 * There is no empty buffer to use.
208 * Allocate a new buffer if:
209 * 1. We can't seek on this file and -b is not in effect; or
210 * 2. We haven't allocated the max buffers for this file yet.
212 if ((autobuf
&& !(ch_flags
& CH_CANSEEK
)) ||
213 (maxbufs
< 0 || ch_nbufs
< maxbufs
))
216 * Allocation failed: turn off autobuf.
221 bp
= bufnode_buf(bn
);
222 BUF_HASH_RM(bn
); /* Remove from old hash chain. */
223 bp
->block
= ch_block
;
225 BUF_HASH_INS(bn
, h
); /* Insert into new hash chain. */
229 pos
= (ch_block
* LBUFSIZE
) + bp
->datasize
;
230 if ((len
= ch_length()) != NULL_POSITION
&& pos
>= len
)
239 * Not at the correct position: must seek.
240 * If input is a pipe, we're in trouble (can't seek on a pipe).
241 * Some data has been lost: just return "?".
243 if (!(ch_flags
& CH_CANSEEK
))
245 if (lseek(ch_file
, (off_t
)pos
, SEEK_SET
) == BAD_LSEEK
)
247 error("seek error", NULL_PARG
);
256 * If we read less than a full block, that's ok.
257 * We use partial block and pick up the rest next time.
259 if (ch_ungotchar
!= -1)
261 bp
->data
[bp
->datasize
] = ch_ungotchar
;
264 } else if (ch_flags
& CH_HELPFILE
)
266 bp
->data
[bp
->datasize
] = helpdata
[ch_fpos
];
270 n
= iread(ch_file
, &bp
->data
[bp
->datasize
],
271 (unsigned int)(LBUFSIZE
- bp
->datasize
));
278 #if MSDOS_COMPILER==WIN32C
282 error("read error", NULL_PARG
);
290 * If we have a log file, write the new data to it.
292 if (!secure
&& logfile
>= 0 && n
> 0)
293 write(logfile
, (char *) &bp
->data
[bp
->datasize
], n
);
300 * If we have read to end of file, set ch_fsize to indicate
301 * the position of the end of file.
309 * We are ignoring EOF.
310 * Wait a while, then try again.
315 parg
.p_string
= wait_message();
318 sleep_ms(2); /* Reduce system load */
322 if (follow_mode
== FOLLOW_NAME
)
324 /* See whether the file's i-number has changed,
325 * or the file has shrunk.
326 * If so, force the file to be closed and
329 POSITION curr_pos
= ch_tell();
330 int r
= stat(get_filename(curr_ifile
), &st
);
331 if (r
== 0 && (st
.st_ino
!= curr_ino
||
332 st
.st_dev
!= curr_dev
||
333 (curr_pos
!= NULL_POSITION
&& st
.st_size
< curr_pos
)))
335 /* screen_trashed=2 causes
336 * make_display to reopen the file. */
348 if (ch_bufhead
!= bn
)
351 * Move the buffer to the head of the buffer chain.
352 * This orders the buffer chain, most- to least-recently used.
358 * Move to head of hash chain too.
364 if (ch_offset
>= bp
->datasize
)
366 * After all that, we still don't have enough data.
367 * Go back and try again.
371 return (bp
->data
[ch_offset
]);
375 * ch_ungetchar is a rather kludgy and limited way to push
376 * a single char onto an input file descriptor.
382 if (c
!= -1 && ch_ungotchar
!= -1)
383 error("ch_ungetchar overrun", NULL_PARG
);
390 * If we haven't read all of standard input into it, do that now.
393 end_logfile(VOID_PARAM
)
395 static int tried
= FALSE
;
399 if (!tried
&& ch_fsize
== NULL_POSITION
)
402 ierror("Finishing logfile", NULL_PARG
);
403 while (ch_forw_get() != EOI
)
414 * Start a log file AFTER less has already been running.
415 * Invoked from the - command; see toggle_option().
416 * Write all the existing buffered data to the log file.
419 sync_logfile(VOID_PARAM
)
427 nblocks
= (ch_fpos
+ LBUFSIZE
- 1) / LBUFSIZE
;
428 for (block
= 0; block
< nblocks
; block
++)
433 bp
= bufnode_buf(bn
);
434 if (bp
->block
== block
)
436 write(logfile
, (char *) bp
->data
, bp
->datasize
);
441 if (!wrote
&& !warned
)
443 error("Warning: log file is incomplete",
453 * Determine if a specific block is currently in one of the buffers.
464 FOR_BUFS_IN_CHAIN(h
, bn
)
466 bp
= bufnode_buf(bn
);
467 if (bp
->block
== block
)
474 * Seek to a specified position in the file.
475 * Return 0 if successful, non-zero if can't seek there.
484 if (thisfile
== NULL
)
488 if (pos
< ch_zero() || (len
!= NULL_POSITION
&& pos
> len
))
491 new_block
= pos
/ LBUFSIZE
;
492 if (!(ch_flags
& CH_CANSEEK
) && pos
!= ch_fpos
&& !buffered(new_block
))
496 while (ch_fpos
< pos
)
498 if (ch_forw_get() == EOI
)
508 ch_block
= new_block
;
509 ch_offset
= pos
% LBUFSIZE
;
514 * Seek to the end of the file.
517 ch_end_seek(VOID_PARAM
)
521 if (thisfile
== NULL
)
524 if (ch_flags
& CH_CANSEEK
)
525 ch_fsize
= filesize(ch_file
);
528 if (len
!= NULL_POSITION
)
529 return (ch_seek(len
));
532 * Do it the slow way: read till end of data.
534 while (ch_forw_get() != EOI
)
541 * Seek to the last position in the file that is currently buffered.
544 ch_end_buffer_seek(VOID_PARAM
)
551 if (thisfile
== NULL
|| (ch_flags
& CH_CANSEEK
))
552 return (ch_end_seek());
557 bp
= bufnode_buf(bn
);
558 buf_pos
= (bp
->block
* LBUFSIZE
) + bp
->datasize
;
559 if (buf_pos
> end_pos
)
563 return (ch_seek(end_pos
));
567 * Seek to the beginning of the file, or as close to it as we can get.
568 * We may not be able to seek there if input is a pipe and the
569 * beginning of the pipe is no longer buffered.
572 ch_beg_seek(VOID_PARAM
)
575 struct bufnode
*firstbn
;
578 * Try a plain ch_seek first.
580 if (ch_seek(ch_zero()) == 0)
584 * Can't get to position 0.
585 * Look thru the buffers for the one closest to position 0.
587 firstbn
= ch_bufhead
;
588 if (firstbn
== END_OF_CHAIN
)
592 if (bufnode_buf(bn
)->block
< bufnode_buf(firstbn
)->block
)
595 ch_block
= bufnode_buf(firstbn
)->block
;
601 * Return the length of the file, if known.
604 ch_length(VOID_PARAM
)
606 if (thisfile
== NULL
)
607 return (NULL_POSITION
);
609 return (NULL_POSITION
);
610 if (ch_flags
& CH_HELPFILE
)
611 return (size_helpdata
);
612 if (ch_flags
& CH_NODATA
)
618 * Return the current position in the file.
623 if (thisfile
== NULL
)
624 return (NULL_POSITION
);
625 return (ch_block
* LBUFSIZE
) + ch_offset
;
629 * Get the current char and post-increment the read pointer.
632 ch_forw_get(VOID_PARAM
)
636 if (thisfile
== NULL
)
641 if (ch_offset
< LBUFSIZE
-1)
652 * Pre-decrement the read pointer and get the new current char.
655 ch_back_get(VOID_PARAM
)
657 if (thisfile
== NULL
)
665 if (!(ch_flags
& CH_CANSEEK
) && !buffered(ch_block
-1))
668 ch_offset
= LBUFSIZE
-1;
674 * Set max amount of buffer space.
675 * bufspace is in units of 1024 bytes. -1 mean no limit.
678 ch_setbufspace(bufspace
)
685 maxbufs
= ((bufspace
* 1024) + LBUFSIZE
-1) / LBUFSIZE
;
692 * Flush (discard) any saved file state, including buffer contents.
699 if (thisfile
== NULL
)
702 if (!(ch_flags
& CH_CANSEEK
))
705 * If input is a pipe, we don't flush buffer contents,
706 * since the contents can't be recovered.
708 ch_fsize
= NULL_POSITION
;
713 * Initialize all the buffers.
717 bufnode_buf(bn
)->block
= -1;
721 * Figure out the size of the file, if we can.
723 ch_fsize
= filesize(ch_file
);
726 * Seek to a known position: the beginning of the file.
729 ch_block
= 0; /* ch_fpos / LBUFSIZE; */
730 ch_offset
= 0; /* ch_fpos % LBUFSIZE; */
734 * This is a kludge to workaround a Linux kernel bug: files in
735 * /proc have a size of 0 according to fstat() but have readable
736 * data. They are sometimes, but not always, seekable.
737 * Force them to be non-seekable here.
742 if (fstatfs(ch_file
, &st
) == 0)
744 if (st
.f_type
== PROC_SUPER_MAGIC
)
746 ch_fsize
= NULL_POSITION
;
747 ch_flags
&= ~CH_CANSEEK
;
753 if (lseek(ch_file
, (off_t
)0, SEEK_SET
) == BAD_LSEEK
)
756 * Warning only; even if the seek fails for some reason,
757 * there's a good chance we're at the beginning anyway.
758 * {{ I think this is bogus reasoning. }}
760 error("seek error to 0", NULL_PARG
);
765 * Allocate a new buffer.
766 * The buffer is added to the tail of the buffer chain.
769 ch_addbuf(VOID_PARAM
)
775 * Allocate and initialize a new buffer and link it
776 * onto the tail of the buffer list.
778 bp
= (struct buf
*) calloc(1, sizeof(struct buf
));
794 init_hashtbl(VOID_PARAM
)
798 for (h
= 0; h
< BUFHASH_SIZE
; h
++)
800 thisfile
->hashtbl
[h
].hnext
= END_OF_HCHAIN(h
);
801 thisfile
->hashtbl
[h
].hprev
= END_OF_HCHAIN(h
);
806 * Delete all buffers for this file.
809 ch_delbufs(VOID_PARAM
)
813 while (ch_bufhead
!= END_OF_CHAIN
)
817 free(bufnode_buf(bn
));
824 * Is it possible to seek on a file descriptor?
832 if (f
== fd0
&& !isatty(fd0
))
835 * In MS-DOS, pipes are seekable. Check for
836 * standard input, and pretend it is not seekable.
841 return (lseek(f
, (off_t
)1, SEEK_SET
) != BAD_LSEEK
);
845 * Force EOF to be at the current read position.
846 * This is used after an ignore_eof read, during which the EOF may change.
849 ch_set_eof(VOID_PARAM
)
851 if (ch_fsize
!= NULL_POSITION
&& ch_fsize
< ch_fpos
)
857 * Initialize file state for a new file.
865 * See if we already have a filestate for this file.
867 thisfile
= (struct filestate
*) get_filestate(curr_ifile
);
868 if (thisfile
== NULL
)
871 * Allocate and initialize a new filestate.
873 thisfile
= (struct filestate
*)
874 ecalloc(1, sizeof(struct filestate
));
875 thisfile
->buflist
.next
= thisfile
->buflist
.prev
= END_OF_CHAIN
;
877 thisfile
->flags
= flags
;
880 thisfile
->offset
= 0;
882 thisfile
->fsize
= NULL_POSITION
;
885 * Try to seek; set CH_CANSEEK if it works.
887 if ((flags
& CH_CANSEEK
) && !seekable(f
))
888 ch_flags
&= ~CH_CANSEEK
;
889 set_filestate(curr_ifile
, (void *) thisfile
);
891 if (thisfile
->file
== -1)
902 int keepstate
= FALSE
;
904 if (thisfile
== NULL
)
907 if ((ch_flags
& (CH_CANSEEK
|CH_POPENED
|CH_HELPFILE
)) && !(ch_flags
& CH_KEEPOPEN
))
910 * We can seek or re-open, so we don't need to keep buffers.
915 if (!(ch_flags
& CH_KEEPOPEN
))
918 * We don't need to keep the file descriptor open
919 * (because we can re-open it.)
920 * But don't really close it if it was opened via popen(),
921 * because pclose() wants to close it.
923 if (!(ch_flags
& (CH_POPENED
|CH_HELPFILE
)))
931 * We don't even need to keep the filestate structure.
935 set_filestate(curr_ifile
, (void *) NULL
);
940 * Return ch_flags for the current file.
943 ch_getflags(VOID_PARAM
)
945 if (thisfile
== NULL
)
952 ch_dump(struct filestate
*fs
)
960 printf(" --no filestate\n");
963 printf(" file %d, flags %x, fpos %x, fsize %x, blk/off %x/%x\n",
964 fs
->file
, fs
->flags
, fs
->fpos
,
965 fs
->fsize
, fs
->block
, fs
->offset
);
966 printf(" %d bufs:\n", fs
->nbufs
);
967 for (bn
= fs
->next
; bn
!= &fs
->buflist
; bn
= bn
->next
)
969 bp
= bufnode_buf(bn
);
970 printf("%x: blk %x, size %x \"",
971 bp
, bp
->block
, bp
->datasize
);
972 for (s
= bp
->data
; s
< bp
->data
+ 30; s
++)
973 if (*s
>= ' ' && *s
< 0x7F)