[help-3dldf] Re: [Re: HELP TO INSTALL 1.2.0.0 RELESASE.]

[Laurence Finston]

------------------

> I tried to sum up  what i did. I tried experiments with make ldf in the
> cweb  directory, in the 3DLDF-.. directory, make install etc. always 
> the same error  appears. 

> 1. I run the 3.4.3 release of gcc :

So you're using a more recent release than I am.  The next time I'm at a
computer where I can do this, I'll  replace `Scanner_Node' with `void*' in the
declarations of the `Pointer_Vector' functions, and move the definitions to a
file processed after `io.web'.  `Scanner_Type' is defined in the latter file.

If I'm doing anything with the `Scanner_Type' object referenced by the
`Scanner_Node' argument in any of these functions, than it's actually correct
(if inconvenient) behavior for the compiler to signal an error.  Perhaps the
GCC developers closed a hole here.  It may have worked before, because the
actual code for the template functions was only generated when they were
called.  In earlier versions of GCC, the template definitions may not have
been checked, or not thoroughly enough to have caught this error.  You're also
using a different GNU/Linux distribution, and it's possible that that makes a
difference.

I don't have a computer of my own, so I have to go to the computer lab to take
care of this.  I plan to go tomorrow, but it closes at 6:00 PM on Saturdays,
so I may not make it.  I do plan to go on Sunday.   I'm sorry to make you
wait.  Of course, if you've understood my explanations, and you know how to do
this, you could just fix your copies of my sources yourself.   I do try to
encourage people to look at the sources.  They may seem intimidating, but it
is possible to understand them.

> 2.1 I downloaded the very last release (about 17:00 today) of the 
> 3DLDFsnp.tar.gz on your site. 

You may find that the Savannah server is faster than the ftp server at the
GWDG.

> 3. Then I went, as you said, in the CWEB directory and made 
> loader.o 

In future, you needn't bother to build `loader.o' and `pspglb.o' separately.

> Makefile:783: attention : écrasement des commandes pour la cible « 
> 3dldf » 
> Makefile:524: attention : anciennes commandes ignorées pour la 
> cible « 3dldf » 

If you reset the environment variable `LANG' in your shell and export it, you
should be able to get error messages in English.  I wish I could speak French,
but unfortunately I can't (yet).  This won't effect other shells that aren't
child processes of the one in question.
On my system, I use `en_us'.  You should be able to find out which locales are
available by invoking `man locale'.

> /bin/sh ./gctng3DL 
> ctangle.w:78: warning: conflicting types for built-in function 'strlen' 
> common.w:1412: warning: conflicting types for built-in function 'strlen' 

`ctangle' is Don Knuth and Silvio Levy's problem.  It's just a warning, so I'm
not going to worry about it.  If I get ambitious, I'll try to find out where
bug reports for `ctangle' are to be sent.

> Is it a bug ? I don't know but I don't manage to run 3DLDF.

If my analysis above is correct, then yes, it is a bug.  I'll take a look at
`pspglb.web' and see if these functions access the `Scanner_Type' object.  

Thanks, and I'll let you know when I've made the changes.  You'll have to tell
me if they fix the problem, though.  

Laurence