Re: DHCP server bug (miscompiled by GCC compiler)

On Mon, Sep 12, 2005 at 04:34:45AM +0300, Raimo Koski wrote:
> So, GPL is irrelevant after all.

What part of the GPL?  Where does it say that Red Hat has to release all
the compilers and tools used to build the binaries?  

> You can use any compiler version, perhaps once released, but no longer 
> available, to build binaries nobody else can reproduce.

One of us misinterpreted what was said.  Basically what happened was
that Red Hat identified a compiler bug that impacted one of their
packages (dhcpd in this case).  What Red Hat did was to fix the compiler
bug and rebuild the package for their customers.  They then put the
fixed compiler on their list of updates - they couldn't release it at
the same time as dhcpd because it didn't have time to go through a full
quality testing cycle.  So they tested it some more (perhaps fixing
other bugs) and then finally released a new version to the open source
community.

> This stinks to high heaven. Relying to compiler bugs to protect 
> commercial firm value!

Nope - their package broke because of a compiler bug.  What would you
have Red Hat do?  Not release dhcpd because their compiler didn't
compile it successfully, or change the package to work around the
compiler bug?  If the compiler is broken, you fix the compiler. 

> Paranoid thought of line would have RH deviced this at will, but I would 
> have more faith in incompetence, bad practices and such.

If I read the thread properly (and I also received some private email
on this), I don't see any incompetence,  bad practise, or Red Hat doing
anything else wrong.  They found a compiler and they fixed it.  It's not
like they changed dhcpd to only work because of a compiler bug.

I did not check the dhcpd source rpm to see if says that a specific
version of the compiler is required or not.  That would have been the
right thing to include in the dhcpd docs.

-- 
Ed Wilts, RHCE
Mounds View, MN, USA
mailto:[email protected]
Member #1, Red Hat Community Ambassador Program