Re: Further reducing special case code for XS structuring

[Patrice Dumas]

On Fri, Apr 05, 2024 at 02:40:41PM +0100, Gavin Smith wrote:
> On Fri, Apr 05, 2024 at 12:20:19AM +0200, Patrice Dumas wrote:
> > I am not sure that no_store is used, but I think that it should be
> > considered completely separately from no_build and the on-demand Perl
> > structures build, as it is a completely different issue.
> 
> parse_texi_piece is declared in Parsetexi.pm as
> "sub parse_texi_piece($$;$$$)" while it is declared in ParserNonXS.pm as
> "sub parse_texi_piece($$;$)", taking two fewer optional arguments.  (I
> believe this hasn't led to an error message from Perl as the function
> signatures are not checked when a sub is called with method call syntax.)
> The $no_store argument doesn't appear to be used anywhere.

Indeed, it is there for the consistency of the API, to have something
similar with parse_texi_piece and parse_texi_line, but it is not
currently used in texi2any or tests code.  I think that it would be
cleaner if it stayed.

> I've committed a change to remove the $no_build parameter but not the
> $no_store parameter.

It seems to me that the changes are no complete, to me the _get_parser_info
$no_build argument should be removed too, and the build_document should
only be called if $no_store is set.  It seems to me that is would be
better for parse_texi_line, it seems to me that if $no_store is not set,
the tree should not be built in _get_parser_info, but in
parse_texi_line, such that the decision to get a tree or only a handler
is possible in future changes of parse_texi_line.

-- 
Pat