[Top][All Lists]
[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]
Re: Warn about character literals not of length one.
From: |
Joel E. Denny |
Subject: |
Re: Warn about character literals not of length one. |
Date: |
Fri, 24 Jul 2009 11:42:11 -0400 (EDT) |
User-agent: |
Alpine 1.00 (DEB 882 2007-12-20) |
On Fri, 24 Jul 2009, Akim Demaille wrote:
> Le 23 juil. 09 à 20:22, Joel E. Denny a écrit :
>
> > I'm surprised to see a comment in scan-gram.l that describes this behavior
> > almost as if it's desired:
> >
> > /* Characters. We don't check there is only one. */
>
> A FIXME might be missing :)
Ah, I didn't think of that.
> > Is there some good reason why we shouldn't check?
>
> I don't see any. Unless POSIX states this behavior, of course.
These are the only relevant statements I've found from Open Group:
A literal shall consist of a single character enclosed in
single-quotes ( '" ). All of the escape sequences supported for
character constants by the ISO C standard shall be supported by
yacc.
So I think we're fine there.
> > I've thought about making it an error instead of a warning because it
> > seems like a severe mistake to miss. However, I worry that could cause
> > trouble for distro maintainers trying to build existing projects.
>
> But we should really make this an error in the future, and I suggest that NEWS
> state this.
Ok, an adjusted patch is below.
I'm feeling confident after your and Eric's reviews, so I pushed to
branch-2.5 and master. Thanks.
>From ac9b0e954b1d3aed514a3bbd363da1514202af0f Mon Sep 17 00:00:00 2001
From: Joel E. Denny <address@hidden>
Date: Fri, 24 Jul 2009 10:29:07 -0400
Subject: [PATCH] Warn about character literals not of length one.
* NEWS (2.5): Document.
* src/scan-gram.l (INITIAL): Remove comment that we don't check
the length.
(SC_ESCAPED_CHARACTER): Warn if length is wrong.
* tests/input.at (Bad character literals): New test group.
diff --git a/NEWS b/NEWS
index 7d401ad..e4bb63f 100644
--- a/NEWS
+++ b/NEWS
@@ -98,6 +98,19 @@ Bison News
about a missing semicolon where it did not before. Future releases of
Bison will cease to append semicolons entirely.
+** Character literals not of length one.
+
+ Previously, Bison quietly converted all character literals to length
+ one. For example, without warning, Bison interpreted the operators in
+ the following grammar to be the same token:
+
+ exp: exp '++'
+ | exp '+' exp
+ ;
+
+ Bison now warns when a character literal is not of length one. In
+ some future release, Bison will report an error instead.
+
* Changes in version 2.4.2 (????-??-??):
** %code is now a permanent feature.
diff --git a/src/scan-gram.l b/src/scan-gram.l
index 60813bb..0c12cba 100644
--- a/src/scan-gram.l
+++ b/src/scan-gram.l
@@ -250,7 +250,7 @@ splice (\\[ \f\t\v]*\n)*
complain_at (*loc, _("invalid identifier: %s"), quote (yytext));
}
- /* Characters. We don't check there is only one. */
+ /* Characters. */
"'" STRING_GROW; token_start = loc->start; BEGIN SC_ESCAPED_CHARACTER;
/* Strings. */
@@ -465,24 +465,40 @@ splice (\\[ \f\t\v]*\n)*
<SC_ESCAPED_CHARACTER>
{
"'"|"\n" {
- if (yytext[0] == '\n')
- unexpected_newline (token_start, "'");
STRING_GROW;
STRING_FINISH;
loc->start = token_start;
val->character = last_string[1];
+ {
+ /* FIXME: Eventually, make these errors. */
+ size_t length = strlen (last_string);
+ if (strlen (last_string) < 3)
+ warn_at (*loc, _("empty character literal"));
+ else if (strlen (last_string) > 3)
+ warn_at (*loc, _("extra characters in character literal"));
+ }
+ if (yytext[0] == '\n')
+ unexpected_newline (token_start, "'");
STRING_FREE;
BEGIN INITIAL;
return CHAR;
}
<<EOF>> {
- unexpected_eof (token_start, "'");
STRING_FINISH;
loc->start = token_start;
- if (strlen (last_string) > 1)
- val->character = last_string[1];
- else
- val->character = last_string[0];
+ {
+ size_t length = strlen (last_string);
+ /* FIXME: Eventually, make these errors. */
+ if (length < 2)
+ warn_at (*loc, _("empty character literal"));
+ else if (length > 2)
+ warn_at (*loc, _("extra characters in character literal"));
+ if (length > 1)
+ val->character = last_string[1];
+ else
+ val->character = last_string[0];
+ }
+ unexpected_eof (token_start, "'");
STRING_FREE;
BEGIN INITIAL;
return CHAR;
diff --git a/tests/input.at b/tests/input.at
index f62a68d..4d5d0bb 100644
--- a/tests/input.at
+++ b/tests/input.at
@@ -1162,3 +1162,65 @@ AT_CHECK_NAMESPACE_ERROR([[::]],
[[namespace reference has a trailing "::"]])
AT_CLEANUP
+
+## ------------------------ ##
+## Bad character literals. ##
+## ------------------------ ##
+
+# Bison used to accept character literals that were empty or contained
+# too many characters.
+
+# FIXME: $ECHO_N and $ECHO_C are not very portable according to the
+# Autoconf manual. Switch to AS_ECHO_N when Autoconf 2.64 is released?
+# Even better, AT_DATA or some variant of AT_DATA may eventually permit
+# a trailing newline. See the threads starting at
+# <http://lists.gnu.org/archive/html/bison-patches/2009-07/msg00019.html>.
+
+AT_SETUP([[Bad character literals]])
+
+AT_DATA([empty.y],
+[[%%
+start: '';
+start: '
+]])
+echo $ECHO_N "start: '$ECHO_C" >> empty.y
+
+AT_BISON_CHECK([empty.y], [1], [],
+[[empty.y:2.8-9: warning: empty character literal
+empty.y:3.8-4.0: warning: empty character literal
+empty.y:3.8-4.0: missing `'' at end of line
+empty.y:4.8: warning: empty character literal
+empty.y:4.8: missing `'' at end of file
+]])
+
+AT_DATA([two.y],
+[[%%
+start: 'ab';
+start: 'ab
+]])
+echo $ECHO_N "start: 'ab$ECHO_C" >> two.y
+
+AT_BISON_CHECK([two.y], [1], [],
+[[two.y:2.8-11: warning: extra characters in character literal
+two.y:3.8-4.0: warning: extra characters in character literal
+two.y:3.8-4.0: missing `'' at end of line
+two.y:4.8-10: warning: extra characters in character literal
+two.y:4.8-10: missing `'' at end of file
+]])
+
+AT_DATA([three.y],
+[[%%
+start: 'abc';
+start: 'abc
+]])
+echo $ECHO_N "start: 'abc$ECHO_C" >> three.y
+
+AT_BISON_CHECK([three.y], [1], [],
+[[three.y:2.8-12: warning: extra characters in character literal
+three.y:3.8-4.0: warning: extra characters in character literal
+three.y:3.8-4.0: missing `'' at end of line
+three.y:4.8-11: warning: extra characters in character literal
+three.y:4.8-11: missing `'' at end of file
+]])
+
+AT_CLEANUP
--
1.5.4.3
- Re: AT_DATA and AS_ECHO_N, (continued)
- Re: AT_DATA and AS_ECHO_N, Paolo Bonzini, 2009/07/29
- Re: AT_DATA and AS_ECHO_N, Joel E. Denny, 2009/07/30
- Re: AT_DATA and AS_ECHO_N, Paolo Bonzini, 2009/07/30
- Re: AT_DATA and AS_ECHO_N, Joel E. Denny, 2009/07/30
- Re: AT_DATA and AS_ECHO_N, Paolo Bonzini, 2009/07/30
- Re: AT_DATA and AS_ECHO_N, Ralf Wildenhues, 2009/07/30
- [PATCH] Clarify comparison of echo, printf, and AS_ECHO*., Paolo Bonzini, 2009/07/30
- Re: [PATCH] Clarify comparison of echo, printf, and AS_ECHO*., Joel E. Denny, 2009/07/31
Re: Warn about character literals not of length one., Joel E. Denny, 2009/07/24
Re: Warn about character literals not of length one., Akim Demaille, 2009/07/24
- Re: Warn about character literals not of length one.,
Joel E. Denny <=