lantz (04/05/83)
There was a report in unix-wizards of a "bug in vi" which caused it to
give a "Too many file names" error, if a file name was given in a :e
command which contained an asterisk, even if only one file name matched.
The same error occurs if the filename starts with a tilde.
This was correctly found by Scott Bradner at Harvard University to be
caused by the fact that csh, which is used by vi to expand the filename,
was printing its prompt in addition to the filename. I had also found
this prompt annoying when other programs, such as Franz Lisp, invoke the
shell to execute one command.
Bradner's correction was not exactly what was called for, however. The csh
has a flag 'prompt' which is true if the shell is supposed to prompt for
input. This flag is set by default, and is reset by certain flags, including
the -c flag, which is the one used by vi and Franz Lisp. The problem is
that THIS FLAG IS NOT CHECKED by the shell before it prints a prompt! The
check goes in the routine process, in sh.c. Diffs follow.
Philip Lantz
duke!ucf-cs!lantz
*** sh.c.old Mon Apr 4 22:33:47 1983
--- sh.c Mon Apr 4 22:33:31 1983
***************
*** 669,673
if (neednote)
pnote();
! if (intty && evalvec == 0) {
mailchk();
--- 669,674 -----
if (neednote)
pnote();
! /* Check of prompt added by PRL, UCF, April 1983. */
! if (intty && prompt && evalvec == 0) {
mailchk();smb (04/06/83)
My experience is that the alleged bug in csh (that it prints the prompt when it shouldn't, thereby affecting file-name expansion from 'vi' and 'ucbmail') isn't a bug; rather, many people unconditionally set 'prompt' in their .cshrc files. It should only be set if it's already set to something: if ($?prompt) set prompt='MyPrompt '
rascal (04/06/83)
I agree with the line: if($?prompt) set prompt='my prompt' You should now all change your .cshrc files. Also, change all your friends. Send a note to msgs. If you are in charge of setting up accounts, remember to put this line in everyone's .cshrc file. Whoever is in charge of a manual, (including vi and lisp), you might want to include a little note. I don't ever want to hear about this problem again. I also don't ever want to hear about shell files that don't work when people have shell variables set (like $D and spell). Thank you all.
chris@pixutl.UUCP (chris) (10/05/84)
There are a couple of bugs in the 'history' command of /bin/csh (and
offspring, such as newcsh):
1) The maximum number of arguments to the history command is set to 2.
2) The validity of the flags is not checked and since the argument pointer
is only incremented when a valid flag is found, using a wrong flag throws
the Cshell in a loop.
-------------
To duplicate, type:
csh
set history = 100
history -h -r 5
history -x
-------------
To fix:
1) In sh.init.c, change the following line:
"history", dohist, 0, 2,
to:
"history", dohist, 0, 4,
2) in sh.hist.c, change the following lines in dohist():
vp++;
while (*vp && *vp[0] == '-') {
if (*vp && eq(*vp, "-h")) {
hflg++;
vp++;
}
if (*vp && eq(*vp, "-r")) {
rflg++;
vp++;
}
}
to:
while (*++vp && **vp == '-') {
while(*++*vp)
switch(**vp) {
case 'h':
hflg++; break;
case 'r':
rflg++; break;
case '-': /* ignore multiple '-'s */
break;
default:
printf("Unknown flag: -%c\n", **vp);
error("Usage: history [-rh] [# of events]");
}
}
-----------------------------------------
--
Chris Bertin : (617) 657-8720 x2318
Pixel Computer Inc. :
260 Fordham Rd. : {allegra|ihnp4|cbosgd|ima|genrad|amd|harvard}\
Wilmington, Ma 01887 : !wjh12!pixel!pixutl!chrisglenn@wacsvax.OZ (Glenn Huxtable) (10/25/84)
A fix was reported for a bug in scanning 'history' command arguments of CSH. I installed this fix, to find another bug. The fix was given as ... >2) The validity of the flags is not checked and since the argument > pointer is only incremented when a valid flag is found, using a > wrong flag throws the Cshell in a loop. >2) in sh.hist.c, change the following lines in dohist(): > vp++; > while (*vp && *vp[0] == '-') { > if (*vp && eq(*vp, "-h")) { > hflg++; > vp++; > } > if (*vp && eq(*vp, "-r")) { > rflg++; > vp++; > } > } >to: > while (*++vp && **vp == '-') { > while(*++*vp) > switch(**vp) { > case 'h': > hflg++; break; > case 'r': > rflg++; break; > case '-': /* ignore multiple '-'s */ > break; > default: > printf("Unknown flag: -%c\n", **vp); > error("Usage: history [-rh] [# of events]"); > } > } The fix introduced another bug, as the line 'while(*++*vp)' modifies the argument pointer in scanning through the argument. Later CSH tries to use this pointer (which now points to the end of the argument) to free the memory used to store the command. Free (malloc) gets an mfree botch and CSH dies. The solution is to use a local pointer 'vp2' say the fix becomes ... => char *vp2; ... while (*++vp && **vp == '-') { => vp2 = *vp; => while(*++vp2) => switch(*vp2) { case 'h': hflg++; break; case 'r': rflg++; break; case '-': /* ignore multiple '-'s */ break; default: => printf("Unknown flag: -%c\n", *vp2); error("Usage: history [-rh] [# of events]"); } } ----------------------------------------------------- Glenn Huxtable Department of Computer Science University of Western Australia USENET: ...decvax!mulga!wacsvax!glenn OZNET: glenn:wacsvax