This is the mail archive of the gcc-patches@gcc.gnu.org mailing list for the GCC project.


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]
Other format: [Raw text]

Re: GCC build of HEAD failed for native with your patch on2004-09-04T15:32:13Z.


On Mon, 2004-09-06 at 18:12 -0700, James E Wilson wrote:
> On Mon, 2004-09-06 at 17:19, Daniel Berlin wrote:
> > I was under the distinct impression that our policy was that the person
> > responsible for uncovering the latent bug is the one who is supposed to
> > fix it, regardless of whether they originally caused it.
> > In fact, I remember reading this numerous times on the mailing list.
> > Is this incorrect?
> > Can somebody please clarify?
> 
> I believe the rule is that if you add a patch that creates a problem,
> then you are responsible for either reverting your patch or fixing the
> bugs it created, even if they are latent.
> 
Okay.

> However, I would argue that I did not add a patch.  I reverted one.  I
> removed an incorrect change added by someone else 5 months ago,
> restoring the exact same behaviour that existed before the incorrect
> patch went in.
> 

I don't necessarily disagree with your assessment. I just wanted to make
sure what i thought was the policy didn't change when i wasn't looking.

> Also, in this case, reverting my patch reversion is not an option, as
> the code really was wrong for 5 months, and we can not go back to the
> broken code.
> 
Right..

> That leaves the issue of what to do about the latent problems that were
> hidden by the broken change that I removed.  I would argue that it is
> wrong to automatically assign responsibility for them to me, because
> that would be counterproductive.  Our goal here is to get the most bugs
> fixed.  But if fixing one person's mistake makes me liable for fixing N
> other mistakes made by M other people in the meantime, then I would be
> less likely to accept responsibility for fixing random bugs in the
> future.  And that would hurt everyone.  Thus in a special case like
> this, I don't think responsibility for the latent bugs should
> automatically fall to me.
> 
I don't personally disagree that fixing bugs should be a different case
than introducing new things that trigger latent bugs, but i do think it
is somewhat of a slippery slope.  Even if the policy stays the way it
is, i certainly believe people fixing bugs should be granted more leeway
(in terms of time to fix, etc) when they trigger latent problems than
people introducing new optimizations.



> In the interest of getting the most bugs fixed, which is my goal here, I
> did look at this problem, write a patch for it, and check it in.  I did
> this as soon as I was able to, but I got stuck with two ppc problems at
> the same time, I have to do ppc testing on a machine I could afford to
> buy with my own money, the rules require so much testing that it took me
> 16 hours to get through the first one, plus the 3-day holiday weekend...

On a completely unrelated note, i'm not sure what speed ppc you have,
but i can at least give you (or any other gcc developmer) access to my
1.6ghz powerbook g4 if they have too slow of a ppc machine.
It's not the fastest thing on the planet, but may be faster than would
you have (i have no idea).  It obviously has gcc, gdb, dejagnu, etc,
installed (I bootstrap gcc on it when at home and i need a ppc test
machine).



Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]