Re: bison 1.35: end of file?

[Akim Demaille]

>>>>> "Torsten" == Torsten Müller <address@hidden> writes:

Torsten> I now took a look onto bison 1.49b (latest version) which
Torsten> produces indeed a compilable output file.

Torsten> But I cannot detect any change in the behaviour in view of my
Torsten> problem.  An existing error-symbol is still _not_ matched if
Torsten> yylex() returns 0.  The bison-parser terminates by calling
Torsten> yyerror() but it does not execute my action below the
Torsten> error-symbol.

So there is probably an error in your grammar.  Can we see it?  What
do the traces reveal about your parser?  Do you have a small test case?