Re: How to get Ada to ?cross the chasm??

[Paul Rubin <no.email@nospam.invalid>]

gautier_niouzes@hotmail.com writes:
> It's corrected in rev. r622:
> https://sourceforge.net/p/unzip-ada/code/622/

This is weird, when I clicked the "download" button it gave me
zipada53.zip.  Is that an obsolete version?  I did an svn checkout and
got the current version (629) which built ok with gprbuild.  It's late
here so I'll test it some other time.

The revision 622 comment says "regressed 'Image to pre-Ada-2012 form,
for pleasing gcc version 5.4.0 20160609".  But I'm using the default
GNAT with Debian 9 and that seems to use gcc6.  Is that supposed to
support Ada 2012?  I thought last time I played with it, I was able to
use Ada 2012 features.

> There is no match because I've programmed the MA part of LZMA from
> scratch.  My version is less sophisticated than the original.

Oh hmm, it would be REALLY useful to have a safe replacement for the
historically buggy 7z and maybe by extension xz.  

> Actually there are probably still obvious bugs waiting around....

If they are subscript overruns that trigger constraint error exceptions
in the Ada code and crash the program, that's still much safer than the
C behaviour of allowing malicious code injection, so it's a huge
improvement.  Going past that calls for using SPARK to show the absence
of subscript and int overflow errors, which might or might not be on
your radar.  Even further (and probably a lot harder) would be SPARK
verification that compressing something and uncompressing it restores
the original input.

>   https://blog.adacore.com/running-american-fuzzy-lop-on-your-ada-code

Oh this is great, I had been wondering whether AFL could be used with
Ada.  It is important.  Thanks.