>Unfortunately, the demoted tools do not only use tiffconf.h.  Some of them 
>rely heavily on libtiff internals.  It wouldn't be a simple process to 
>separate the tools from the rest of libtiff.

I didn't realize that. I think that is worth fixing.
The utilities can't be examples of best use of libtiff if they access libtiff 
internals that normal applications can't use.
Possibly other users of libtiff are forced into accessing the same libtiff 
internals, and it might help libtiff users if everything required to implement 
the utilities was part of the stable public api.
Is it more complicated than trying to build the utilities with just the public 
header files to see what breaks, and then moving some declarations or wrappers 
to access them to a public header?

>Fork libtiff

How about setting up the utilities as a git repo that builds with libtiff as a 
submodule?
The utilities would be able to reference a specific commit of libtiff, so they 
could access libtiff internals safely.

>I have been trying to rally people interested in doing the work to fix the 
>security issues.

I could probably handle an occasional request with a cve and example tiff that 
is reproducible on Linux x86_64.

William

________________________________
From: Lee Howard <fax...@howardsilvan.com>
Sent: Sunday, February 4, 2024 4:15 PM
To: William Bader <williamba...@hotmail.com>; Bob Friesenhahn 
<bfrie...@simple.dallas.tx.us>; Rob Tillaart <rob.tilla...@gmail.com>
Cc: Patrice Fournier via Tiff <tiff@lists.osgeo.org>
Subject: Re: [Tiff] www.libtiff.org is restored

On 2/4/24 8:11 AM, William Bader via Tiff wrote:
>If the utilities are again maintained by the libtiff project, they
need to be in a separate repository, with its own build process

Shouldn't the build for the utilities be simpler than the build for libtiff 
because the utilities only use tiffconf.h instead of generating it?

The utilities should be best use examples for libtiff, and something might be 
wrong if they end up needing a massively complicated build process.


Unfortunately, the demoted tools do not only use tiffconf.h.  Some of them rely 
heavily on libtiff internals.  It wouldn't be a simple process to separate the 
tools from the rest of libtiff.


I understand and hear the opinions about having the tools be in a separate 
repository maintained by separate people, etc., but the tools can't really be 
ripped out of the library source and compile at all.  To maintain the tools 
would necessarily require maintaining the library.  The tools are integrated 
into the library source.


The only paths forward that I can visualize would be:


1) Fix the security issues in the tools, commit to maintaining them long term, 
and restore the tools to the package as they were previously.  This is what I 
had understood was acceptable to Even and Sulau.  (And I'm a bit discouraged 
about slight backpedaling I'm hearing in this thread.)


2) Fork libtiff - one with the tools, and one without.  (And the one with the 
tools would necessarily duplicate the one without.)  Distributions would either 
need to choose between the two or package both in ways in which they wouldn't 
conflict, e.g., libtiff45 and libtiff.


Again, I had heard that #1 was acceptable, and I have been trying to rally 
people interested in doing the work to fix the security issues.  I'll do all of 
that work, myself, if I ultimately have to (once I can get some priority work 
out of the way first).  I find it frustrating to hear comments here from dev 
team members suggesting that #1 is now not acceptable.


I am in no way interested in #2.  However, if #1 will never, ever happen then I 
am faced with no other practical option.


Thanks,


Lee.

_______________________________________________
Tiff mailing list
Tiff@lists.osgeo.org
https://lists.osgeo.org/mailman/listinfo/tiff

Reply via email to