Re: LYNX-DEV New code for testing

[Klaus Weide]

On Tue, 22 Apr 1997, Laura Eaves wrote:

> I built it on linux as the last time I tried bsdi it didn't configure...
> (don't know if that's fixed yet)
> 
> Here are some comments:
> I configured it as followed:
>       configure --with-screen=slang --prefix=$HOME
> It configured and built fine, but when I ran it it looked for lynx.cfg
> in /usr/local/lib instead of $HOME/lib.  I think if configure is going
> to intall things in $prefix/{bin,lib,man} then it should also configure
> the software to build that way so you don't have to edit userdefs.h.

Seems reasonable that the location should be set consistently 
*as a default*, but it could not replace the requirement
that the installer have a look at the configuration files and edit
them if appropriate.  There are just to many things that can be
set differently, and moving them all from the files to the ./configure
command line (which then would manipulate the files) somehow doesn't
seem to make sense.

> Also, when you unzip the source, it puts everything in lynx2-7-1/
> It'd be nice if it unpacked in lynx3.0 or something so it wouldn't
> clobber my old source.

I think for others that is a feature.  Since I sometimes make .zip's that
only contain the new or changed fiiles, for those it is a requirement.
You can simply create a (sub)directory somewhere else and start from there
if you have a lynx2-7-1/ which you don;'t want to change.
 
> Now as for the new parsing stuff,
> I tried it on http://www.atitech.ca/
> and it worked great; all problems fixed.
> 
> However, I then tried it on
>       http://www.voicenet.com/~leaves/file.html
> (one of the pages with the form parsing problem) and it completely
> screwed up.  SELECT options were printed as normal text.
> I tried it on
>       http://www.voicenet.com/~leaves/f.html
> (same file with the bad html fixed) and it still complained about
> bad html and dumped all the select options.
>
> Anyway, the most important parsing fix for my purposes is the form bug fix.
> So if/when that's working I'll consider trying it again...

Thanks for trying it.

You should get much better results on those after changing just one
character (actually, one *bit*) in one line in one file...

In WWW/Library/Implementation/HTMLDTD.c, locate the line starting with

#define T_SELECT        

and change the last digit on that line from '1' to '0'
(which it should have been in the first place)...
That will tell the {start,end}_element functions in SGML.c that
the </SELECT> end tag cannot be legally omitted, so any old
unexpected tag (like </OPTION> in your case) will not cause closing
premature of the <SELECT>.

If you have better ideas for the parser, please pass them on...

   Klaus

;
; To UNSUBSCRIBE:  Send a mail message to address@hidden
;                  with "unsubscribe lynx-dev" (without the
;                  quotation marks) on a line by itself.
;