Re: parsing template?

[Ben Pfaff]

On Sat, May 30, 2015 at 03:28:27PM -0500, Alan Mead wrote:
> I filed a bug because CLUSTER hangs on unknown subcommands.  It turns
> out that (for reasons I don't fully understand) the main parsing loop
> never terminates when an unsupported or invalid subcommand is used.  I
> don't understand how the parsing works with all the lex_*() functions. 
> It looks like when any unexpected subcommand is present that the rest of
> the command is ignored and the "lex_token (lexer) != T_ENDCMD" condition
> is never met.
> 
> Is there a working template of how parsing is supposed to occur?
> 
> I can fix the issue by adding a final else at the end of parser:
> 
> $ diff quick-cluster.c quick-cluster.c.original
> 579,584d578
> <       else
> <         {
> <         lex_error (lexer, _("Invalid or unsupported subcommand"));
> <         goto error;
> <       }
> <
> 
> Is there a reason not to do this? One reason is that it will break any
> SPSS syntax that uses unsupported subcommands but that seems better than
> having PSPP hang.

It's just a bug.  I fixed it with a patch similar to your suggestion.