[ros-dev] [ros-diffs] [pschweitzer] 49662: [NTOSKRNL] [HAL] Disable INIT_FUN

Timo Kreuzer timo.kreuzer at web.de
Sat Nov 20 21:37:24 UTC 2010


Am 20.11.2010 18:13, Ros Arm wrote:
> Hi,
>
> It appears your arguments are because you don't seem to understand what you have done wrong.
Let me add my portion here as well.

Despite the fact that the scientific method doesn't seem to apply to
reactos coding (if you have evaluated all possibilities and all of them
are impossible except one that is unlikely, then ... one of the
impossible possibilities is responsible), you would need to take
everything into account. One of these things to be taken into account is
human failability, and yes, even your own failability ;-) Even the
possibility that you didn't take everything into account. Well that
makes the task quite impossible to solve.

Another thing you have to take into account is that we have a tool
called pefixup, which messes with INIT sections. I don't know if you
were aware of that, but before blaming the compiler, I would look at
that tool.

Also following your argumentation, that it can only be a compiler bug,
as the kernel doesn't do anything with INIT sections yet, this bug would
obviously affect other modules as well, even if we didn't notice it (you
called it a subtle bug) and we might end up with having it being
triggered on the kernel by some completely unrelated commit, making it
almost impossible to find out what happened. So not reverting the kernel
(and win32k) would be ignoring this fact.

I remember spending hours, trying to find out, why merging the header
branch back to trunk caused a testbot failure. It turned out that adding
some function prototypes to ddk headers triggered the bug. It was
completely unrelated to the content, simply adding more stuff into a
header broke boot. Impossible? Still it's a fact. It's always better to
not make quick assumptions about what is possible and what not.

If you asked me, whether I would prefer reverting this commit or having
testbot being broken for even longer, I would go for reverting. Always.
Instantly. No matter who, what and where. And I'm certain that quite a
number of developers and testers would agree. Reverting something that
doesn't work, doesn't mean ignoring the problem, it only means that
developers and testers are tired of having to deal with broken shit on a
daily basis.

You seemed to have been aware that something broke for some people.
Still you switched to completely unrelated things and had other people
deal with it. I don't complain about that, but either fix it yourself,
or live with other people's "fixes".

I don't care for anyone's epeen. But I care for having a working
testbot. Pierre succeeded in bringing it back, you failed.

Finally: instead of ranting, you might as well have politely asked to
bring back the kernel's INIT_FUNCTION. Telling somebody that his
"decision is breaindead" and that he doesn't "get the scientific method"
doesn't help anyone except maybe your ego.

I just want to teach you :-p

Regards,
Timo




More information about the Ros-dev mailing list