\chapter{Warning and Error Messages}
\index{Warning messages}\index{Error messages}\index{Messages}\index{Messages}

AnaGram provides diagnostic messages in the \agwindow{Warnings} window
when it finds constructs in your grammar which might be incorrect, or
when it finds that it cannot continue its analysis of your grammar.
AnaGram gives up only if it encounters a syntax error or finds that
you have not specified a \index{Grammar token}\index{Token}grammar
token.  Otherwise, it continues its analysis.  If AnaGram succeeds in
analyzing your grammar, it will be in a position to build a parser for
you, if you should want one.

When AnaGram finds a syntax error in its input it attempts to continue
scanning for syntax errors but it does no further analysis of the
input.

When AnaGram identifies a problem while it is scanning your syntax
file, it precedes the diagnostic with the line number and column
number at which it found the problem.  It does not provide line and
column numbers for those problems it discovers when it is actually
analyzing the grammar.

As an aid to debugging, for all warnings developed while scanning the
syntax file, AnaGram locks the cursor in the syntax file window to the
cursor in the Warnings window.  This means that when you move the
cursor bar in the Warnings window to a warning message that has a line
number and column number, the cursor in the syntax file window is
moved to the corresponding line and column.

\section{Alphabetic Listing of Warning Messages}

In this list of warning messages, each message is classified according
to whether it is a warning or an error and according to when it
occurs.  Errors prevent AnaGram from building a parser; warnings do
not.  Messages may occur at program initialization, while scanning the
syntax file, while analyzing the grammar, or while building the
parser.  Messages issued while scanning the syntax file are always
accompanied by the line and column number in the syntax file where the
problem is recognized.

% XXX I would like to change ``warning, analysis'' to ``analysis
% warning'' for all these.

\dualheading{Both error token resynch and auto resynch specified}
{warning, analysis}

Your grammar defines an \index{Error token}\index{Token}\agparam{error
token} and also requests automatic resynchronization. AnaGram will
ignore the request for automatic
\index{Resynchronization}resynchronization and will provide error
token \index{Resynchronization}resynchronization. If you named a token
\agcode{error} but do not wish error token resynchronization, you can
either rename \agcode{error}, or explicitly specify the error token to be
something you don't use in your grammar:

\begin{indentingcode}{0.4in}
{}[ error token = not used ]
\end{indentingcode}

\dualheading{Conflicts resolved by precedence rules}
{warning, analysis}

Your grammar has conflicts which AnaGram has resolved by applying the
precedence rules you supplied.  These conflicts are listed in the
\index{Resolved Conflicts}\index{Window}\agwindow{Resolved Conflicts}
window, and are not listed in the
\index{Conflicts}\index{Window}\agwindow{Conflicts} window.

\dualheading{Duplicate production}
{warning, scan}

A \index{Production}production has been duplicated.  You will have a
number of reduce-reduce \index{Conflicts}conflicts as a consequence.
Eliminate the duplicate, and the conflicts it caused will go away.

\dualheading{Empty keyword string}
{warning, scan}

\index{Keywords}
Keyword strings must contain at least one character.  If you wish a
null match, use a null production instead.

\dualheading{eof token not defined}
{warning, analysis}

\index{Eof token}\index{Token}
You have requested either error token or automatic resynchronization
and you have not defined an end of file token. The resynchronization
procedure will not work correctly at end of file.

\dualheading{Error defining T\textit{nnn}: \textit{token-representation}}
{warning, analysis}

Errors were encountered while attempting to evaluate the
\index{Character set}character set for token number \textit{nnn}.
This warning is always generated in addition to more detailed warnings
that were made when the actual errors were encountered.

\dualheading{Event driven parser cannot use pointer input}
{error, build}

\index{Parser}
\index{Configuration switches}
\index{Event driven}
\index{Pointer input}
If you are going to use \agparam{pointer input}, you should not
specify your parser as \agparam{event driven}.
Conversely, if you really want an event driven parser, you cannot specify
\agparam{pointer input}.  See Appendix A, Configuration Parameters for
details on the \agparam{pointer input} and \agparam{event driven}
switches.

\dualheading{Excessive recursion}
{error, scan}

An internal \index{Stack}stack in AnaGram overflowed because of the
complexity of an expression in your grammar.  Simplify your grammar by
using definition statements to name subexpressions.

\dualheading{Fatal syntax errors}
{error, analysis}

AnaGram could not complete its analysis of your syntax file because of
syntax errors.

\dualheading{Grammar is ambiguous. See Conflicts Window}
{warning, analysis}

Your grammar is ambiguous.  See Chapter 7 for suggestions on how to
deal with the ambiguities.

\dualheading{Inappropriate value}
{warning, scan}

The value assigned to a configuration parameter is not appropriate.
The value of the parameter is unchanged.  Check the definition of the
parameter in Appendix A, Configuration Parameters.

\dualheading{Input scan aborted}
{error, scan}

Because of the previous errors, AnaGram was unable to finish scanning
your syntax file.

\dualheading{Insufficient memory}
{warning, scan or analysis}

Not enough memory is available to continue parsing or analyzing the
current syntax file.  If you have been running AnaGram for some time,
you might succeed in analyzing your grammar if you restart AnaGram.
Otherwise, you need more memory, or you need to make your grammar
somewhat less complex.

\dualheading{Internal error: ...}
{error, any phase}

This error results from internal consistency checks that should never
fail if AnaGram is
working properly.  If you are running AnaGram on Windows 95 or Windows
98, it is
probably not advisable to continue running AnaGram after receiving
this warning without rebooting your system.  Sometimes when you are
testing buggy programs using an operating system that does not have
complete memory protection, a program may inadvertently alter part of
the system's memory.  If this happens, problems may snowball.  In this
situation the problem will not recur after you reboot your system.  If
the problem recurs on a freshly booted system, it may be a problem in
AnaGram which should be fixed. Please note the complete error message
and file a bug report via \url{http://www.parsifalsoft.com}.
\index{Internal error}\index{Errors}
A copy of the syntax file which causes the problem
would be of great assistance in fixing it.

\dualheading{Keyword anomalies found}
{warning, analysis}

See the
\index{Anomaly}\index{Keyword Anomalies}\agwindow{Keyword Anomalies}
table for a list of the anomalies found in your syntax file.
See Chapter 7 for a treatment of keyword anomalies.

\dualheading{Missing \textit{syntactic-element-1} in
\textit{syntactic-element-2}}
{error, scan}

AnaGram expects to see an instance of \textit{syntactic element 1} at
this location in an instance of \textit{syntactic element 2}.  It
cannot reliably continue parsing its input.  Therefore, it limits
further analysis to scanning for syntax errors.

\dualheading{Missing production, T\textit{nnn}: \textit{token-name}}
{error, scan}

\index{Production}
Token number \textit{nnn}, a nonterminal token, appears to be defined
recursively, but there is no initial production to get the recursion
started.  Check your grammar closely.

\dualheading{Missing reduction procedure, R\textit{nnn}}
{warning, scan}

\index{Reduction procedure}
A reduction parameter has been defined in this rule, number
\textit{nnn}, but there is no reduction procedure to use it.  This
warning is provided in case the lack of a reduction procedure is an
oversight.

\dualheading{Multiple definitions for T\textit{nnn}: \textit{token-name}}
{warning, analysis}

\index{Definitions}
Token number \textit{nnn} has been defined as a character set and has
also been defined as a nonterminal token.  It can't be both.

\dualheading{Negative character codes in pointer mode}
{warning, analysis}

\index{Character codes}
You have defined negative character codes and you are using pointer
input.  If you are using the default definition for pointer type you
will be reading unsigned characters, so that your parser will never
see the negative codes you have defined.

% I wish there were an easy way to shut this up if the pointer type is
% signed.

\dualheading{No grammar token specified}
{error, analysis}

\index{Grammar token}\index{Token}
You have not specified a grammar token.
Edit your syntax file to specify one.

\dualheading{No productions in syntax file}
{error, analysis}

AnaGram did not find any productions at all in the syntax file.  Do
you have the right file?

\dualheading{No such parameter}
{warning, scan}

AnaGram does not recognize the name of the configuration parameter you
have tried to set.  Check the spelling of the parameter you wish to
set.  See Appendix A, Configuration Parameters.

\dualheading{Not a character set, T\textit{nnn}: \textit{token-name}}
{warning, analysis}

Token number \textit{nnn} has been used on the left side of a
production and in a character set expression defining some other
token.  AnaGram will use an empty set in place of token number
\textit{nnn} in the character set expression.  You will get another
warning, Error defining token, when AnaGram finishes its evaluation of
the character set.

\dualheading{Nothing reduces T\textit{nnn} -$>$ R\textit{mmm}}
{warning, analysis}

Your grammar does not specify any input to follow an instance of this
rule, rule number \textit{mmm}.  In all probability, your grammar does
not have any explicit end of file indicator.  If you do not have any
conflicts with token T000, then you can get along without an end of
file indicator.  Otherwise, you should modify your grammar so that it
recognizes end of file.

\dualheading{Null character in keyword string}
{warning, scan}

\index{Keywords}
A keyword string cannot have an embedded null character.  If you must
allow for a null you will have to rewrite your grammar rule.  For
example, instead of

\begin{indentingcode}{0.4in}
widget
  -> "abc{\bs}0def"
\end{indentingcode}
write
\begin{indentingcode}{0.4in}
widget
  -> "abc", 0, "def"
\end{indentingcode}

If you have defined a symbol as a keyword string, you will have to
rewrite the definition as a production. Suppose, for example, you had
written

\begin{indentingcode}{0.4in}
marker = "abc{\bs}0def"
\end{indentingcode}

You would have to rewrite this as

\begin{indentingcode}{0.4in}
marker
  -> "abc", 0, "def"
\end{indentingcode}

\dualheading{Parameter \textit{name} has type void}
{warning, scan}

The specified parameter for a reduction procedure attempts to take the
value of a token whose type has been specified as void.

\dualheading{Range bounds out of order}
{warning, scan}

\index{Character range}\index{Range}
AnaGram interprets 'z-a', for example, as the letters from a to z, but
provides a warning in case the unusual order was a typo.

\dualheading{Recursive definition of char set: \textit{name}}
{warning, analysis}

\index{Character sets}
Character sets cannot be defined recursively.

\dualheading{Redefinition of \textit{name}}
{warning, analysis}

There is a previous definition statement for this variable.  AnaGram
ignores the new definition.

\dualheading{Redefinition of grammar token}
{warning, scan}

\index{Grammar token}\index{Token}
AnaGram simply discards the old definition of the grammar, or goal,
token and uses the new one.  The last definition in the syntax file
wins.  Check your syntax file to be sure you have the grammar token
you want.

\dualheading{Redefinition of token, T\textit{nnn}: \textit{name}}
{warning, scan}

This variable has already been used as a nonterminal token, that is,
it has been used as the left side of a production.  AnaGram will
ignore this definition statement.

\dualheading{Suspicious production for \textit{token name}}
{warning, scan}

This production has the form

\begin{indentingcode}{0.4in}
x -> x
\end{indentingcode}

You probably have a typo somewhere.  This production causes a conflict
in your grammar.  AnaGram leaves this production in your grammar, but
if you build a parser, it will never succeed in recognizing this
production.

\dualheading{Syntax Analysis Aborted}
{error, analysis}

Because of the previous errors, AnaGram was unable to complete its
analysis of your syntax file.

\dualheading{Token not used, T\textit{nnn}: \textit{token name}}
{warning, analysis}

AnaGram doesn't mind, but warns you just in case there was some sort
of oversight.

\dualheading{Type Redefinition of T\textit{nnn}: \textit{token name}}
{warning, analysis}

A previous statement has defined a different data type for token
number \textit{nnn}.  This definition will override the previous one.
If you intend to use different type definitions, you should use
extreme caution and check the generated code to verify that your
reduction procedures are getting the values you intended.

\dualheading{Undefined symbol: \textit{name}}
{warning, analysis}

\index{Symbol}
AnaGram encountered this undefined symbol while evaluating a character
set.  The next warning in the Warnings window identifies the
particular token AnaGram was trying to evaluate.

\dualheading{Undefined token T\textit{nnn}: \textit{name}}
{warning, analysis}

\index{Token}
This name has been used as though it represented a token, but there is
no definition of it as a terminal nor does any production define it as
a nonterminal.

\dualheading{Unexpected \textit{syntactic-element-1} in
\textit{syntactic-element-2}}
{error, scan}

AnaGram does not expect to see an instance of
\textit{syntactic-element-1} at this location in an instance of
\textit{syntactic-element-2}.
It cannot reliably continue parsing its input.  Therefore, it limits
further analysis to scanning for syntax errors.  If this error is not
the result of some prior error, you should correct your syntax file.
Remember that this error could just as well result from something
missing as well as from something extraneous.

\dualheading{Void token, \textit{token-name}, used as parameter}
{warning, scan}

\index{Token}
The token being declared as void has previously been used as a
parameter for a reduction procedure.  Your C compiler will complain
when it tries to compile the call to the reduction procedure.