eval + format - \n = pp_ctl.c assertion #7009
Comments
From rob@exitexchange.comThis is a bug report for perl from rob@exitexchange.com, created a format with eval, that does NOT include a trailing \n, Assertion fpc <= fops + maxops failed: file "pp_ctl.c", line 3731 at (eval 1) line 5. Test case: my @v = ('k'); # eval "format STDOUT = \n@\n\@v\n"; # does NOT trigger bug write; Flags: Site configuration information for perl v5.8.2: Configured by rob at Mon Nov 17 11:00:52 PST 2003. Summary of my perl5 (revision 5.0 version 8 subversion 2) configuration: Locally applied patches: @INC for perl v5.8.2: Environment for perl v5.8.2: Complete configuration data for perl v5.8.2: Author='' |
From @iabynOn Tue, Dec 30, 2003 at 01:26:59AM -0000, rob@exitexchange.com (via RT) wrote:
Ah, mea cupla! Fixed by patch #22005 -- Change 22005 by davem@davem-percy on 2003/12/30 15:07:50 Subject: Re: [perl #24774] eval + format - \n = pp_ctl.c assertion Affected files ... ... //depot/perl/pp_ctl.c#377 edit Differences ... ==== //depot/perl/pp_ctl.c#377 (text) ==== @@ -3580,7 +3580,7 @@ /* estimate the buffer size needed */ ==== //depot/perl/t/op/write.t#31 (xtext) ==== @@ -5,7 +5,7 @@ -print "1..49\n"; my +{ -# 14..49: scary format testing from Merijn H. Brand -my $test = 14; if ( |
|
@iabyn - Status changed from 'open' to 'resolved' |
From wolfgang.laun@alcatel.atOn Tue, 30 Dec 2003 15:30:37 +0000 davem@fdisolutions.com wrote:
Festina lente ;-) Yes, it fixes the assertion failure and all that may come from an bash$ cat fmtbug.pl The lone semicolon is sneaked into the string to be eval'ed, and no check A patch against toke.c (5.8.2) appears to fix this: bash$ ./perl fmtbug.pl This breaks the new test 14 in t/op/write.t, so this is patched again. Finally I'd like to draw your attention to the fact that some NULL Regards Inline Patch--- toke.c.old Fri Jan 2 08:50:37 2004
+++ toke.c Fri Jan 2 10:11:31 2004
@@ -2424,8 +2424,12 @@
if (!PL_rsfp) {
PL_last_uni = 0;
PL_last_lop = 0;
- if (PL_lex_brackets)
- yyerror("Missing right curly or square bracket");
+ if (PL_lex_brackets) {
+ if (PL_lex_formbrack)
+ yyerror("Format not terminated");
+ else
+ yyerror("Missing right curly or square bracket");
+ }
DEBUG_T( { PerlIO_printf(Perl_debug_log,
"### Tokener got EOF\n");
} );
@@ -7577,6 +7581,7 @@
register char *t;
SV *stuff = newSVpvn("",0);
bool needargs = FALSE;
+ bool eofmt = FALSE;
while (!needargs) {
if (*s == '.' || *s == /*{*/'}') {
@@ -7586,8 +7591,10 @@
#else
for (t = s+1;SPACE_OR_TAB(*t) || *t == '\r'; t++) ;
#endif
- if (*t == '\n' || t == PL_bufend)
+ if (*t == '\n' || t == PL_bufend) {
+ eofmt = TRUE;
break;
+ }
}
if (PL_in_eval && !PL_rsfp) {
eol = strchr(s,'\n');
@@ -7627,7 +7634,6 @@
PL_last_lop = PL_last_uni = Nullch;
if (!s) {
s = PL_bufptr;
- yyerror("Format not terminated");
break;
}
}
@@ -7656,7 +7662,8 @@
}
else {
SvREFCNT_dec(stuff);
- PL_lex_formbrack = 0;
+ if (eofmt)
+ PL_lex_formbrack = 0;
PL_bufptr = s;
}
return s;
--- t/op/write.t.old Fri Jan 2 11:11:28 2004
+++ t/op/write.t Fri Jan 2 10:53:50 2004
@@ -302,12 +302,10 @@
{
# Bug #24774 format without trailing \n failed assertion
+ # but this must not compile because we'd get a ';' into the format
my @v = ('k');
eval "format OUT14 = \n@\n\@v";
- open(OUT14, '>Op_write.tmp') || die "Can't create Op_write.tmp";
- write(OUT14);
- close OUT14 or die "Could not close: $!";
- print "ok 14\n";
+ print $@ ? "ok 14\n" : "not ok 14\n";
}
####################################### |
From @TuxOn Fri 02 Jan 2004 11:31, LAUN Wolfgang <wolfgang.laun@alcatel.at> wrote:
Wolfgang, can I please have your comments on this: On Mon 27 Oct 2003 19:36, "H.Merijn Brand" <h.m.brand@hccnet.nl> wrote:
-- |
From wolfgang.laun@alcatel.atHi Merijn, just a quick answer:
Well, even then you'd want to keep escapes.
These things happen in pp_ctl.c, and not just to ESC either. It seems THIS PATCH IS UNTESTED! I'll be back.
Inline Patch--- pp_ctl.c.new Fri Jan 2 12:45:32 2004
+++ pp_ctl.c Fri Jan 2 12:50:13 2004
@@ -621,23 +621,15 @@
*t = '\0';
sv_catpvn_utf8_upgrade(PL_formtarget, s, arg, nsv);
for (; t < SvEND(PL_formtarget); t++) {
-#ifdef EBCDIC
- int ch = *t++ = *s++;
- if (iscntrl(ch))
-#else
- if (!(*t & ~31))
-#endif
+ int ch = *t = *s++;
+ if (isSPACE(ch))
*t = ' ';
}
break;
}
while (arg--) {
-#ifdef EBCDIC
int ch = *t++ = *s++;
- if (iscntrl(ch))
-#else
- if ( !((*t++ = *s++) & ~31) )
-#endif
+ if (isSPACE(ch))
t[-1] = ' ';
}
break; |
From @iabynOn Fri, Jan 02, 2004 at 11:31:46AM +0100, LAUN Wolfgang wrote:
Nil illegitimo carborundum?
Thanks, applied as 22055.
My personal take on this (not that I ever use formats!) is that since Dave. -- |
Migrated from rt.perl.org#24774 (status was 'resolved')
Searchable as RT24774$
The text was updated successfully, but these errors were encountered: