Re: [Texmacs-dev] [Announce] nogencc-0.8

[Joris van der Hoeven]

On Tue, 7 May 2002, David Allouche wrote:

> New in this release:
> --------------------
> 
>   New experimental make target ALL_IN_ONE (fails).

What is this supposed to mean?
Please give a short explanation of the different targets.

>   New experimental make target STATIC_AGGREGATE (fails).

Please fix these problems. Before that such important problems
have been fixed, we cannot consider nogencc as sufficiently stable.

>   Added ffid4emaner script which does the opposite of rename4diff.

Thanks.

>   Replaced new refcounting by old one, operator-> are no longer const
>   correct.
> 
>   Fixed partial rebuild problem with AGGREGATE target.

Thanks.

> About STATIC_AGGREGATE:
> -----------------------
> 
>   You may notice that the STATIC_AGGREGATE target fails differently than 
> the official distribution. There, the error message is:
> 
> /lib/ld-linux.so.2(*IND*+0x0): multiple definition of `__xstat64'
> /lib/ld-linux.so.2(.text+0xc810): first defined here
> /lib/ld-linux.so.2(*IND*+0x0): multiple definition of `__fxstat64'
> /lib/ld-linux.so.2(.text+0xc9b0): first defined here
> /usr/bin/ld: BFD 2.12.90.0.1 20020307 Debian/GNU Linux assertion fail 
> ../../bfd/elf-strtab.c:262
> collect2: ld returned 1 exit status
> 
>   To me, it looks like a stupid error, but I do not understand what is 
> going on.

You probably did not specify the linker options in the right order/way.
I also don't understand why /lib/ld-linux.so.2 is being linked;
this is a dynamic library, while you are building a static version!

> Now, what?
> ----------
> 
> About the static link problem in the official distribution, I WAS using 
> guile 1.4. Until someone brings me new information, I will consider this 
> issue out of my competence.

Well, it is not in my competence either. I consider this to be
a big problem on which you should try hard to solve.
My experience tells me that it is very unlikely that an average user
will help you on this problem in a reasonable delay. We cannot make
TeXmacs instable until a friendly user ultimately shows up and
solves the problem.

> Unless you require me to remove the modifications I made for the sake of 
> separate compilation (addition of includes and declarations, and creation 
> of a few header files), I will consider this version stable. Removing 
> these changes would require at least a couple of days of dull and tedious 
> work.

The includes should not be a big problem;
I am more concerned about the actual code.

> I would like you to validate nogencc so I can synch it with 1.0.0.4, and 
> then move to more interesting work. Now that the new refcounting has been 
> removed, the conversion script should be easy to understand. That way you 
> can avoid making a brute force diff on the converted code (which is going 
> to be long and uninteresting).

You should first solve the problems mentioned above.
You might post a message on some mailing lists
in order to get help on the linking problems.
I have no time to help you on this issue.

When the above problems will be fixed, I will proceed as follows:

1. I try to compile the nogencc distribution on some other platforms
   and let you fix possible problems.

2. Sync with TeXmacs-1.0.0.5 and freeze the C++ code:
   I will not make any changes in the C++ code until
   the nogencc problem has been solved.

3. I make a complete diff (I do not trust automatic rewriting) and
   study all differences. If I find non trivial modifications in
   the code, then I will ask you to fix these.

4. In the meantime you should adapt TeXmacs-misc too;
   I will again make a complete diff for this too.

5. Validate nogencc in TeXmacs-1.0.0.6 or
   temporarily invalidate nogencc if there are too many problems.

6. We will wait a few weeks/months for feedback on possible problems.
   When I will be sufficiently confident that the change did not
   break anything, then we will rediscuss the other changes that
   you proposed.