838 lines
24 KiB
C
838 lines
24 KiB
C
/*-
|
|
* Copyright (c) 1992, 1993, 1994
|
|
* The Regents of the University of California. All rights reserved.
|
|
*
|
|
* Redistribution and use in source and binary forms, with or without
|
|
* modification, are permitted provided that the following conditions
|
|
* are met:
|
|
* 1. Redistributions of source code must retain the above copyright
|
|
* notice, this list of conditions and the following disclaimer.
|
|
* 2. Redistributions in binary form must reproduce the above copyright
|
|
* notice, this list of conditions and the following disclaimer in the
|
|
* documentation and/or other materials provided with the distribution.
|
|
* 3. All advertising materials mentioning features or use of this software
|
|
* must display the following acknowledgement:
|
|
* This product includes software developed by the University of
|
|
* California, Berkeley and its contributors.
|
|
* 4. Neither the name of the University nor the names of its contributors
|
|
* may be used to endorse or promote products derived from this software
|
|
* without specific prior written permission.
|
|
*
|
|
* THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
|
|
* ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
|
|
* IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
|
|
* ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
|
|
* FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
|
|
* DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
|
|
* OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
|
|
* HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
|
|
* LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
|
|
* OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
|
|
* SUCH DAMAGE.
|
|
*/
|
|
|
|
#ifndef lint
|
|
static char sccsid[] = "@(#)exf.c 8.97 (Berkeley) 8/17/94";
|
|
#endif /* not lint */
|
|
|
|
#include <sys/param.h>
|
|
#include <sys/queue.h>
|
|
#include <sys/stat.h>
|
|
#include <sys/time.h>
|
|
|
|
/*
|
|
* We include <sys/file.h>, because the flock(2) and open(2) #defines
|
|
* were found there on historical systems. We also include <fcntl.h>
|
|
* because the open(2) #defines are found there on newer systems.
|
|
*/
|
|
#include <sys/file.h>
|
|
|
|
#include <bitstring.h>
|
|
#include <errno.h>
|
|
#include <fcntl.h>
|
|
#include <limits.h>
|
|
#include <signal.h>
|
|
#include <stdio.h>
|
|
#include <stdlib.h>
|
|
#include <string.h>
|
|
#include <termios.h>
|
|
#include <unistd.h>
|
|
|
|
#include "compat.h"
|
|
#include <db.h>
|
|
#include <regex.h>
|
|
#include <pathnames.h>
|
|
|
|
#include "vi.h"
|
|
#include "excmd.h"
|
|
|
|
/*
|
|
* file_add --
|
|
* Insert a file name into the FREF list, if it doesn't already
|
|
* appear in it.
|
|
*
|
|
* !!!
|
|
* The "if it doesn't already appear" changes vi's semantics slightly. If
|
|
* you do a "vi foo bar", and then execute "next bar baz", the edit of bar
|
|
* will reflect the line/column of the previous edit session. Historic nvi
|
|
* did not do this. The change is a logical extension of the change where
|
|
* vi now remembers the last location in any file that it has ever edited,
|
|
* not just the previously edited file.
|
|
*/
|
|
FREF *
|
|
file_add(sp, name)
|
|
SCR *sp;
|
|
CHAR_T *name;
|
|
{
|
|
FREF *frp;
|
|
|
|
/*
|
|
* Return it if it already exists. Note that we test against the
|
|
* user's name, whatever that happens to be, including if it's a
|
|
* temporary file.
|
|
*/
|
|
if (name != NULL)
|
|
for (frp = sp->frefq.cqh_first;
|
|
frp != (FREF *)&sp->frefq; frp = frp->q.cqe_next)
|
|
if (!strcmp(frp->name, name))
|
|
return (frp);
|
|
|
|
/* Allocate and initialize the FREF structure. */
|
|
CALLOC(sp, frp, FREF *, 1, sizeof(FREF));
|
|
if (frp == NULL)
|
|
return (NULL);
|
|
|
|
/*
|
|
* If no file name specified, or if the file name is a request
|
|
* for something temporary, file_init() will allocate the file
|
|
* name. Temporary files are always ignored.
|
|
*/
|
|
if (name != NULL && strcmp(name, TEMPORARY_FILE_STRING) &&
|
|
(frp->name = strdup(name)) == NULL) {
|
|
FREE(frp, sizeof(FREF));
|
|
msgq(sp, M_SYSERR, NULL);
|
|
return (NULL);
|
|
}
|
|
|
|
/* Append into the chain of file names. */
|
|
CIRCLEQ_INSERT_TAIL(&sp->frefq, frp, q);
|
|
|
|
return (frp);
|
|
}
|
|
|
|
/*
|
|
* file_init --
|
|
* Start editing a file, based on the FREF structure. If successsful,
|
|
* let go of any previous file. Don't release the previous file until
|
|
* absolutely sure we have the new one.
|
|
*/
|
|
int
|
|
file_init(sp, frp, rcv_name, force)
|
|
SCR *sp;
|
|
FREF *frp;
|
|
char *rcv_name;
|
|
int force;
|
|
{
|
|
EXF *ep;
|
|
RECNOINFO oinfo;
|
|
struct stat sb;
|
|
size_t psize;
|
|
int fd;
|
|
char *oname, tname[MAXPATHLEN];
|
|
|
|
/*
|
|
* If the file is a recovery file, let the recovery code handle it.
|
|
* Clear the FR_RECOVER flag first -- the recovery code does set up,
|
|
* and then calls us! If the recovery call fails, it's probably
|
|
* because the named file doesn't exist. So, move boldly forward,
|
|
* presuming that there's an error message the user will get to see.
|
|
*/
|
|
if (F_ISSET(frp, FR_RECOVER)) {
|
|
F_CLR(frp, FR_RECOVER);
|
|
return (rcv_read(sp, frp));
|
|
}
|
|
|
|
/*
|
|
* Required FRP initialization; the only flag we keep is the
|
|
* cursor information.
|
|
*/
|
|
F_CLR(frp, ~FR_CURSORSET);
|
|
|
|
/*
|
|
* Required EXF initialization:
|
|
* Flush the line caches.
|
|
* Default recover mail file fd to -1.
|
|
* Set initial EXF flag bits.
|
|
*/
|
|
CALLOC_RET(sp, ep, EXF *, 1, sizeof(EXF));
|
|
ep->c_lno = ep->c_nlines = OOBLNO;
|
|
ep->rcv_fd = ep->fcntl_fd = -1;
|
|
LIST_INIT(&ep->marks);
|
|
F_SET(ep, F_FIRSTMODIFY);
|
|
|
|
/*
|
|
* If no name or backing file, create a backing temporary file, saving
|
|
* the temp file name so we can later unlink it. If the user never
|
|
* named this file, copy the temporary file name to the real name (we
|
|
* display that until the user renames it).
|
|
*/
|
|
if ((oname = frp->name) == NULL || stat(oname, &sb)) {
|
|
(void)snprintf(tname,
|
|
sizeof(tname), "%s/vi.XXXXXX", O_STR(sp, O_DIRECTORY));
|
|
if ((fd = mkstemp(tname)) == -1) {
|
|
msgq(sp, M_SYSERR, "Temporary file");
|
|
goto err;
|
|
}
|
|
(void)close(fd);
|
|
|
|
if (frp->name == NULL)
|
|
F_SET(frp, FR_TMPFILE);
|
|
if ((frp->tname = strdup(tname)) == NULL ||
|
|
frp->name == NULL && (frp->name = strdup(tname)) == NULL) {
|
|
if (frp->tname != NULL)
|
|
free(frp->tname);
|
|
msgq(sp, M_SYSERR, NULL);
|
|
(void)unlink(tname);
|
|
goto err;
|
|
}
|
|
oname = frp->tname;
|
|
psize = 4 * 1024;
|
|
F_SET(frp, FR_NEWFILE);
|
|
} else {
|
|
/*
|
|
* Try to keep it at 10 pages or less per file. This
|
|
* isn't friendly on a loaded machine, btw.
|
|
*/
|
|
if (sb.st_size < 40 * 1024)
|
|
psize = 4 * 1024;
|
|
else if (sb.st_size < 320 * 1024)
|
|
psize = 32 * 1024;
|
|
else
|
|
psize = 64 * 1024;
|
|
|
|
ep->mtime = sb.st_mtime;
|
|
|
|
if (!S_ISREG(sb.st_mode))
|
|
msgq(sp, M_ERR,
|
|
"Warning: %s is not a regular file", oname);
|
|
}
|
|
|
|
/* Set up recovery. */
|
|
memset(&oinfo, 0, sizeof(RECNOINFO));
|
|
oinfo.bval = '\n'; /* Always set. */
|
|
oinfo.psize = psize;
|
|
oinfo.flags = F_ISSET(sp->gp, G_SNAPSHOT) ? R_SNAPSHOT : 0;
|
|
if (rcv_name == NULL) {
|
|
if (!rcv_tmp(sp, ep, frp->name))
|
|
oinfo.bfname = ep->rcv_path;
|
|
} else {
|
|
if ((ep->rcv_path = strdup(rcv_name)) == NULL) {
|
|
msgq(sp, M_SYSERR, NULL);
|
|
goto err;
|
|
}
|
|
oinfo.bfname = ep->rcv_path;
|
|
F_SET(ep, F_MODIFIED);
|
|
}
|
|
|
|
/* Open a db structure. */
|
|
if ((ep->db = dbopen(rcv_name == NULL ? oname : NULL,
|
|
O_NONBLOCK | O_RDONLY, DEFFILEMODE, DB_RECNO, &oinfo)) == NULL) {
|
|
msgq(sp, M_SYSERR, rcv_name == NULL ? oname : rcv_name);
|
|
goto err;
|
|
}
|
|
|
|
/*
|
|
* Do the remaining things that can cause failure of the new file,
|
|
* mark and logging initialization.
|
|
*/
|
|
if (mark_init(sp, ep) || log_init(sp, ep))
|
|
goto err;
|
|
|
|
/*
|
|
* Close the previous file; if that fails, close the new one and
|
|
* run for the border.
|
|
*
|
|
* !!!
|
|
* There's a nasty special case. If the user edits a temporary file,
|
|
* and then does an ":e! %", we need to re-initialize the backing
|
|
* file, but we can't change the name. (It's worse -- we're dealing
|
|
* with *names* here, we can't even detect that it happened.) Set a
|
|
* flag so that the file_end routine ignores the backing information
|
|
* of the old file if it happens to be the same as the new one.
|
|
*
|
|
* !!!
|
|
* Side-effect: after the call to file_end(), sp->frp may be NULL.
|
|
*/
|
|
F_SET(frp, FR_DONTDELETE);
|
|
if (sp->ep != NULL && file_end(sp, sp->ep, force)) {
|
|
(void)file_end(sp, ep, 1);
|
|
goto err;
|
|
}
|
|
F_CLR(frp, FR_DONTDELETE);
|
|
|
|
/*
|
|
* Lock the file; if it's a recovery file, it should already be
|
|
* locked. Note, we acquire the lock after the previous file
|
|
* has been ended, so that we don't get an "already locked" error
|
|
* for ":edit!".
|
|
*
|
|
* XXX
|
|
* While the user can't interrupt us between the open and here,
|
|
* there's a race between the dbopen() and the lock. Not much
|
|
* we can do about it.
|
|
*
|
|
* XXX
|
|
* We don't make a big deal of not being able to lock the file. As
|
|
* locking rarely works over NFS, and often fails if the file was
|
|
* mmap(2)'d, it's far too common to do anything like print an error
|
|
* message, let alone make the file readonly. At some future time,
|
|
* when locking is a little more reliable, this should change to be
|
|
* an error.
|
|
*/
|
|
if (rcv_name == NULL)
|
|
switch (file_lock(oname,
|
|
&ep->fcntl_fd, ep->db->fd(ep->db), 0)) {
|
|
case LOCK_FAILED:
|
|
F_SET(frp, FR_UNLOCKED);
|
|
break;
|
|
case LOCK_UNAVAIL:
|
|
msgq(sp, M_INFO,
|
|
"%s already locked, session is read-only", oname);
|
|
F_SET(frp, FR_RDONLY);
|
|
break;
|
|
case LOCK_SUCCESS:
|
|
break;
|
|
}
|
|
|
|
/*
|
|
* The -R flag, or doing a "set readonly" during a session causes
|
|
* all files edited during the session (using an edit command, or
|
|
* even using tags) to be marked read-only. Changing the file name
|
|
* (see ex/ex_file.c), clears this flag.
|
|
*
|
|
* Otherwise, try and figure out if a file is readonly. This is a
|
|
* dangerous thing to do. The kernel is the only arbiter of whether
|
|
* or not a file is writeable, and the best that a user program can
|
|
* do is guess. Obvious loopholes are files that are on a file system
|
|
* mounted readonly (access catches this one on a few systems), or
|
|
* alternate protection mechanisms, ACL's for example, that we can't
|
|
* portably check. Lots of fun, and only here because users whined.
|
|
*
|
|
* !!!
|
|
* Historic vi displayed the readonly message if none of the file
|
|
* write bits were set, or if an an access(2) call on the path
|
|
* failed. This seems reasonable. If the file is mode 444, root
|
|
* users may want to know that the owner of the file did not expect
|
|
* it to be written.
|
|
*
|
|
* Historic vi set the readonly bit if no write bits were set for
|
|
* a file, even if the access call would have succeeded. This makes
|
|
* the superuser force the write even when vi expects that it will
|
|
* succeed. I'm less supportive of this semantic, but it's historic
|
|
* practice and the conservative approach to vi'ing files as root.
|
|
*
|
|
* It would be nice if there was some way to update this when the user
|
|
* does a "^Z; chmod ...". The problem is that we'd first have to
|
|
* distinguish between readonly bits set because of file permissions
|
|
* and those set for other reasons. That's not too hard, but deciding
|
|
* when to reevaluate the permissions is trickier. An alternative
|
|
* might be to turn off the readonly bit if the user forces a write
|
|
* and it succeeds.
|
|
*
|
|
* XXX
|
|
* Access(2) doesn't consider the effective uid/gid values. This
|
|
* probably isn't a problem for vi when it's running standalone.
|
|
*/
|
|
if (O_ISSET(sp, O_READONLY) || !F_ISSET(frp, FR_NEWFILE) &&
|
|
(!(sb.st_mode & (S_IWUSR | S_IWGRP | S_IWOTH)) ||
|
|
access(frp->name, W_OK)))
|
|
F_SET(frp, FR_RDONLY);
|
|
|
|
/*
|
|
* Set the alternate file name to be the file we've just discarded.
|
|
*
|
|
* !!!
|
|
* If the current file was a temporary file, the call to file_end()
|
|
* unlinked it and free'd the name. So, there is no previous file,
|
|
* and there is no alternate file name. This matches historical
|
|
* practice, although in historical vi it could only happen as the
|
|
* result of the initial command, i.e. if vi was executed without a
|
|
* file name.
|
|
*/
|
|
set_alt_name(sp, sp->frp == NULL ? NULL : sp->frp->name);
|
|
|
|
/*
|
|
* Switch...
|
|
*
|
|
* !!!
|
|
* Note, because the EXF structure is examine at interrupt time,
|
|
* the underlying DB structures have to be consistent as soon as
|
|
* it's assigned to an SCR structure.
|
|
*/
|
|
++ep->refcnt;
|
|
sp->ep = ep;
|
|
sp->frp = frp;
|
|
return (0);
|
|
|
|
err: if (frp->name != NULL) {
|
|
free(frp->name);
|
|
frp->name = NULL;
|
|
}
|
|
if (frp->tname != NULL) {
|
|
(void)unlink(frp->tname);
|
|
free(frp->tname);
|
|
frp->tname = NULL;
|
|
}
|
|
if (ep->rcv_path != NULL) {
|
|
free(ep->rcv_path);
|
|
ep->rcv_path = NULL;
|
|
}
|
|
if (ep->db != NULL)
|
|
(void)ep->db->close(ep->db);
|
|
FREE(ep, sizeof(EXF));
|
|
return (1);
|
|
}
|
|
|
|
/*
|
|
* file_end --
|
|
* Stop editing a file.
|
|
*/
|
|
int
|
|
file_end(sp, ep, force)
|
|
SCR *sp;
|
|
EXF *ep;
|
|
int force;
|
|
{
|
|
FREF *frp;
|
|
|
|
/*
|
|
* Clean up the FREF structure.
|
|
*
|
|
* Save the cursor location.
|
|
*
|
|
* XXX
|
|
* It would be cleaner to do this somewhere else, but by the time
|
|
* ex or vi knows that we're changing files it's already happened.
|
|
*/
|
|
frp = sp->frp;
|
|
frp->lno = sp->lno;
|
|
frp->cno = sp->cno;
|
|
F_SET(frp, FR_CURSORSET);
|
|
|
|
/*
|
|
* We may no longer need the temporary backing file, so clean it
|
|
* up. We don't need the FREF structure either, if the file was
|
|
* never named, so lose it.
|
|
*
|
|
* !!!
|
|
* Re: FR_DONTDELETE, see the comment above in file_init().
|
|
*/
|
|
if (!F_ISSET(frp, FR_DONTDELETE) && frp->tname != NULL) {
|
|
if (unlink(frp->tname))
|
|
msgq(sp, M_SYSERR, "%s: remove", frp->tname);
|
|
free(frp->tname);
|
|
frp->tname = NULL;
|
|
if (F_ISSET(frp, FR_TMPFILE)) {
|
|
CIRCLEQ_REMOVE(&sp->frefq, frp, q);
|
|
free(frp->name);
|
|
free(frp);
|
|
}
|
|
sp->frp = NULL;
|
|
}
|
|
|
|
/*
|
|
* Clean up the EXF structure.
|
|
*
|
|
* sp->ep MAY NOT BE THE SAME AS THE ARGUMENT ep, SO DON'T USE IT!
|
|
*
|
|
* If multiply referenced, just decrement the count and return.
|
|
*/
|
|
if (--ep->refcnt != 0)
|
|
return (0);
|
|
|
|
/* Close the db structure. */
|
|
if (ep->db->close != NULL && ep->db->close(ep->db) && !force) {
|
|
msgq(sp, M_ERR, "%s: close: %s", frp->name, strerror(errno));
|
|
++ep->refcnt;
|
|
return (1);
|
|
}
|
|
|
|
/* COMMITTED TO THE CLOSE. THERE'S NO GOING BACK... */
|
|
|
|
/* Stop logging. */
|
|
(void)log_end(sp, ep);
|
|
|
|
/* Free up any marks. */
|
|
(void)mark_end(sp, ep);
|
|
|
|
/*
|
|
* Delete recovery files, close the open descriptor, free recovery
|
|
* memory. See recover.c for a description of the protocol.
|
|
*
|
|
* XXX
|
|
* Unlink backup file first, we can detect that the recovery file
|
|
* doesn't reference anything when the user tries to recover it.
|
|
* There's a race, here, obviously, but it's fairly small.
|
|
*/
|
|
if (!F_ISSET(ep, F_RCV_NORM)) {
|
|
if (ep->rcv_path != NULL && unlink(ep->rcv_path))
|
|
msgq(sp, M_ERR,
|
|
"%s: remove: %s", ep->rcv_path, strerror(errno));
|
|
if (ep->rcv_mpath != NULL && unlink(ep->rcv_mpath))
|
|
msgq(sp, M_ERR,
|
|
"%s: remove: %s", ep->rcv_mpath, strerror(errno));
|
|
}
|
|
if (ep->fcntl_fd != -1)
|
|
(void)close(ep->fcntl_fd);
|
|
if (ep->rcv_fd != -1)
|
|
(void)close(ep->rcv_fd);
|
|
if (ep->rcv_path != NULL)
|
|
free(ep->rcv_path);
|
|
if (ep->rcv_mpath != NULL)
|
|
free(ep->rcv_mpath);
|
|
|
|
FREE(ep, sizeof(EXF));
|
|
return (0);
|
|
}
|
|
|
|
/*
|
|
* file_write --
|
|
* Write the file to disk. Historic vi had fairly convoluted
|
|
* semantics for whether or not writes would happen. That's
|
|
* why all the flags.
|
|
*/
|
|
int
|
|
file_write(sp, ep, fm, tm, name, flags)
|
|
SCR *sp;
|
|
EXF *ep;
|
|
MARK *fm, *tm;
|
|
char *name;
|
|
int flags;
|
|
{
|
|
struct stat sb;
|
|
FILE *fp;
|
|
FREF *frp;
|
|
MARK from, to;
|
|
u_long nlno, nch;
|
|
int btear, fd, noname, oflags, rval;
|
|
char *msg;
|
|
|
|
frp = sp->frp;
|
|
if (name == NULL) {
|
|
noname = 1;
|
|
name = frp->name;
|
|
} else
|
|
noname = 0;
|
|
|
|
/* Can't write files marked read-only, unless forced. */
|
|
if (!LF_ISSET(FS_FORCE) && noname && F_ISSET(frp, FR_RDONLY)) {
|
|
if (LF_ISSET(FS_POSSIBLE))
|
|
msgq(sp, M_ERR,
|
|
"Read-only file, not written; use ! to override");
|
|
else
|
|
msgq(sp, M_ERR, "Read-only file, not written");
|
|
return (1);
|
|
}
|
|
|
|
/* If not forced, not appending, and "writeany" not set ... */
|
|
if (!LF_ISSET(FS_FORCE | FS_APPEND) && !O_ISSET(sp, O_WRITEANY)) {
|
|
/* Don't overwrite anything but the original file. */
|
|
if ((!noname || F_ISSET(frp, FR_NAMECHANGE)) &&
|
|
!stat(name, &sb)) {
|
|
if (LF_ISSET(FS_POSSIBLE))
|
|
msgq(sp, M_ERR,
|
|
"%s exists, not written; use ! to override", name);
|
|
else
|
|
msgq(sp, M_ERR, "%s exists, not written", name);
|
|
return (1);
|
|
}
|
|
|
|
/*
|
|
* Don't write part of any existing file. Only test for the
|
|
* original file, the previous test catches anything else.
|
|
*/
|
|
if (!LF_ISSET(FS_ALL) && noname && !stat(name, &sb)) {
|
|
if (LF_ISSET(FS_POSSIBLE))
|
|
msgq(sp, M_ERR,
|
|
"Use ! to write a partial file");
|
|
else
|
|
msgq(sp, M_ERR, "Partial file, not written");
|
|
return (1);
|
|
}
|
|
}
|
|
|
|
/*
|
|
* Figure out if the file already exists -- if it doesn't, we display
|
|
* the "new file" message. The stat might not be necessary, but we
|
|
* just repeat it because it's easier than hacking the previous tests.
|
|
* The information is only used for the user message and modification
|
|
* time test, so we can ignore the obvious race condition.
|
|
*
|
|
* If the user is overwriting a file other than the original file, and
|
|
* O_WRITEANY was what got us here (neither force nor append was set),
|
|
* display the "existing file" messsage. Since the FR_NAMECHANGE flag
|
|
* is cleared on a successful write, the message only appears once when
|
|
* the user changes a file name. This is historic practice.
|
|
*
|
|
* One final test. If we're not forcing or appending, and we have a
|
|
* saved modification time, stop the user if it's been written since
|
|
* we last edited or wrote it, and make them force it.
|
|
*/
|
|
if (stat(name, &sb))
|
|
msg = ": new file";
|
|
else {
|
|
msg = "";
|
|
if (!LF_ISSET(FS_FORCE | FS_APPEND)) {
|
|
if (ep->mtime && sb.st_mtime > ep->mtime) {
|
|
msgq(sp, M_ERR,
|
|
"%s: file modified more recently than this copy%s",
|
|
name, LF_ISSET(FS_POSSIBLE) ?
|
|
"; use ! to override" : "");
|
|
return (1);
|
|
}
|
|
if (!noname || F_ISSET(frp, FR_NAMECHANGE))
|
|
msg = ": existing file";
|
|
}
|
|
}
|
|
|
|
/* Set flags to either append or truncate. */
|
|
oflags = O_CREAT | O_WRONLY;
|
|
if (LF_ISSET(FS_APPEND))
|
|
oflags |= O_APPEND;
|
|
else
|
|
oflags |= O_TRUNC;
|
|
|
|
/* Open the file. */
|
|
if ((fd = open(name, oflags, DEFFILEMODE)) < 0) {
|
|
msgq(sp, M_SYSERR, name);
|
|
return (1);
|
|
}
|
|
|
|
/* Use stdio for buffering. */
|
|
if ((fp = fdopen(fd, "w")) == NULL) {
|
|
(void)close(fd);
|
|
msgq(sp, M_SYSERR, name);
|
|
return (1);
|
|
}
|
|
|
|
/* Build fake addresses, if necessary. */
|
|
if (fm == NULL) {
|
|
from.lno = 1;
|
|
from.cno = 0;
|
|
fm = &from;
|
|
if (file_lline(sp, ep, &to.lno))
|
|
return (1);
|
|
to.cno = 0;
|
|
tm = &to;
|
|
}
|
|
|
|
/* Turn on the busy message. */
|
|
btear = F_ISSET(sp, S_EXSILENT) ? 0 : !busy_on(sp, "Writing...");
|
|
rval = ex_writefp(sp, ep, name, fp, fm, tm, &nlno, &nch);
|
|
if (btear)
|
|
busy_off(sp);
|
|
|
|
/*
|
|
* Save the new last modification time -- even if the write fails
|
|
* we re-init the time. That way the user can clean up the disk
|
|
* and rewrite without having to force it.
|
|
*/
|
|
ep->mtime = stat(name, &sb) ? 0 : sb.st_mtime;
|
|
|
|
/* If the write failed, complain loudly. */
|
|
if (rval) {
|
|
if (!LF_ISSET(FS_APPEND))
|
|
msgq(sp, M_ERR, "%s: WARNING: file truncated!", name);
|
|
return (1);
|
|
}
|
|
|
|
/*
|
|
* Once we've actually written the file, it doesn't matter that the
|
|
* file name was changed -- if it was, we've already whacked it.
|
|
*/
|
|
F_CLR(frp, FR_NAMECHANGE);
|
|
|
|
/*
|
|
* If wrote the entire file clear the modified bit. If the file was
|
|
* written back to the original file name and the file is a temporary,
|
|
* set the "no exit" bit. This permits the user to write the file and
|
|
* use it in the context of the file system, but still keeps them from
|
|
* losing their changes by exiting.
|
|
*/
|
|
if (LF_ISSET(FS_ALL)) {
|
|
F_CLR(ep, F_MODIFIED);
|
|
if (F_ISSET(frp, FR_TMPFILE))
|
|
if (noname)
|
|
F_SET(frp, FR_TMPEXIT);
|
|
else
|
|
F_CLR(frp, FR_TMPEXIT);
|
|
}
|
|
|
|
msgq(sp, M_INFO, "%s%s%s: %lu line%s, %lu characters",
|
|
INTERRUPTED(sp) ? "Interrupted write: " : "",
|
|
name, msg, nlno, nlno == 1 ? "" : "s", nch);
|
|
|
|
return (0);
|
|
}
|
|
|
|
/*
|
|
* file_m1 --
|
|
* First modification check routine. The :next, :prev, :rewind, :tag,
|
|
* :tagpush, :tagpop, ^^ modifications check.
|
|
*/
|
|
int
|
|
file_m1(sp, ep, force, flags)
|
|
SCR *sp;
|
|
EXF *ep;
|
|
int force, flags;
|
|
{
|
|
/*
|
|
* If the file has been modified, we'll want to write it back or
|
|
* fail. If autowrite is set, we'll write it back automatically,
|
|
* unless force is also set. Otherwise, we fail unless forced or
|
|
* there's another open screen on this file.
|
|
*/
|
|
if (F_ISSET(ep, F_MODIFIED))
|
|
if (O_ISSET(sp, O_AUTOWRITE)) {
|
|
if (!force &&
|
|
file_write(sp, ep, NULL, NULL, NULL, flags))
|
|
return (1);
|
|
} else if (ep->refcnt <= 1 && !force) {
|
|
msgq(sp, M_ERR,
|
|
"File modified since last complete write; write or use %s to override",
|
|
LF_ISSET(FS_POSSIBLE) ? "!" : ":edit!");
|
|
return (1);
|
|
}
|
|
|
|
return (file_m3(sp, ep, force));
|
|
}
|
|
|
|
/*
|
|
* file_m2 --
|
|
* Second modification check routine. The :edit, :quit, :recover
|
|
* modifications check.
|
|
*/
|
|
int
|
|
file_m2(sp, ep, force)
|
|
SCR *sp;
|
|
EXF *ep;
|
|
int force;
|
|
{
|
|
/*
|
|
* If the file has been modified, we'll want to fail, unless forced
|
|
* or there's another open screen on this file.
|
|
*/
|
|
if (F_ISSET(ep, F_MODIFIED) && ep->refcnt <= 1 && !force) {
|
|
msgq(sp, M_ERR,
|
|
"File modified since last complete write; write or use ! to override");
|
|
return (1);
|
|
}
|
|
|
|
return (file_m3(sp, ep, force));
|
|
}
|
|
|
|
/*
|
|
* file_m3 --
|
|
* Third modification check routine.
|
|
*/
|
|
int
|
|
file_m3(sp, ep, force)
|
|
SCR *sp;
|
|
EXF *ep;
|
|
int force;
|
|
{
|
|
/*
|
|
* Don't exit while in a temporary files if the file was ever modified.
|
|
* The problem is that if the user does a ":wq", we write and quit,
|
|
* unlinking the temporary file. Not what the user had in mind at all.
|
|
* We permit writing to temporary files, so that user maps using file
|
|
* system names work with temporary files.
|
|
*/
|
|
if (F_ISSET(sp->frp, FR_TMPEXIT) && ep->refcnt <= 1 && !force) {
|
|
msgq(sp, M_ERR,
|
|
"File is a temporary; exit will discard modifications");
|
|
return (1);
|
|
}
|
|
return (0);
|
|
}
|
|
|
|
/*
|
|
* file_lock --
|
|
* Get an exclusive lock on a file.
|
|
*
|
|
* XXX
|
|
* The default locking is flock(2) style, not fcntl(2). The latter is
|
|
* known to fail badly on some systems, and its only advantage is that
|
|
* it occasionally works over NFS.
|
|
*
|
|
* Furthermore, the semantics of fcntl(2) are wrong. The problems are
|
|
* two-fold: you can't close any file descriptor associated with the file
|
|
* without losing all of the locks, and you can't get an exclusive lock
|
|
* unless you have the file open for writing. Someone ought to be shot,
|
|
* but it's probably too late, they may already have reproduced. To get
|
|
* around these problems, nvi opens the files for writing when it can and
|
|
* acquires a second file descriptor when it can't. The recovery files
|
|
* are examples of the former, they're always opened for writing. The DB
|
|
* files can't be opened for writing because the semantics of DB are that
|
|
* files opened for writing are flushed back to disk when the DB session
|
|
* is ended. So, in that case we have to acquire an extra file descriptor.
|
|
*/
|
|
enum lockt
|
|
file_lock(name, fdp, fd, iswrite)
|
|
char *name;
|
|
int fd, *fdp, iswrite;
|
|
{
|
|
#if !defined(USE_FCNTL) && defined(LOCK_EX)
|
|
/* Hurrah! We've got flock(2). */
|
|
/*
|
|
* !!!
|
|
* We need to distinguish a lock not being available for the file
|
|
* from the file system not supporting locking. Flock is documented
|
|
* as returning EWOULDBLOCK; add EAGAIN for good measure, and assume
|
|
* they are the former. There's no portable way to do this.
|
|
*/
|
|
errno = 0;
|
|
return (flock(fd, LOCK_EX | LOCK_NB) ?
|
|
errno == EAGAIN || errno == EWOULDBLOCK ?
|
|
LOCK_UNAVAIL : LOCK_FAILED : LOCK_SUCCESS);
|
|
|
|
#else /* Gag me. We've got fcntl(2). */
|
|
struct flock arg;
|
|
int didopen, sverrno;
|
|
|
|
arg.l_type = F_WRLCK;
|
|
arg.l_whence = 0; /* SEEK_SET */
|
|
arg.l_start = arg.l_len = 0;
|
|
arg.l_pid = 0;
|
|
|
|
/* If the file descriptor isn't opened for writing, it must fail. */
|
|
if (!iswrite) {
|
|
if (name == NULL || fdp == NULL)
|
|
return (LOCK_FAILED);
|
|
if ((fd = open(name, O_RDWR, 0)) == -1)
|
|
return (LOCK_FAILED);
|
|
*fdp = fd;
|
|
didopen = 1;
|
|
}
|
|
|
|
errno = 0;
|
|
if (!fcntl(fd, F_SETLK, &arg))
|
|
return (LOCK_SUCCESS);
|
|
if (didopen) {
|
|
sverrno = errno;
|
|
(void)close(fd);
|
|
errno = sverrno;
|
|
}
|
|
|
|
/*
|
|
* !!!
|
|
* We need to distinguish a lock not being available for the file
|
|
* from the file system not supporting locking. Fcntl is documented
|
|
* as returning EACCESS and EAGAIN; add EWOULDBLOCK for good measure,
|
|
* and assume they are the former. There's no portable way to do this.
|
|
*/
|
|
return (errno == EACCES || errno == EAGAIN || errno == EWOULDBLOCK ?
|
|
LOCK_UNAVAIL : LOCK_FAILED);
|
|
#endif
|
|
}
|