This is the mail archive of the 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]

Re: Loop unroll fixes

>I did not voice the ultimatum only based on the actions and
>messages archived.  I made the statement after months of private action
>and discussions.

I will accept your statement that you sent private mail that was ignored.
I will not accept your statement that this justifies your action.  There
is no public evidence that a patch was ignored.  Therefore, there is no public
evidence to justify your action.  I will not accept secret communications
as justifcation for your action, hence your action was unwarranted.

In my opinion, your over reliance on private mail was a mistake here.
First of all, you create a single point of failure.  If only one person
is given a chance to review a patch, then it is much less likely to be
reviewed than a patch sent to a mailing list.  That one person might be
busy, or might be on vacation.  Also, it has the effect of hiding the
problem.  You knew that the patch was overdue for a review, but the gcc
developer community did not know, because the patch was not discussed in
public.  If the patch had been discussed in public, other people would
have volunteered to help, and the patch would not have been ignored for
two months.

>I have reviewed the patch.  I and others have tested the patch.

That is good to know.  However, it should have been stated publicly before
you took drastic action.  As far as I know, it was not.  There was only
a statement from the contributor claiming that the patch had been tested.

>	GCC SC members and other developers with GCC commit privileges
>need to take action when the maintainers do not respond to repeated,
>polite prompting.

I can accept that.  But I can not accept use of secret communications
as justification for such actions.  There is too much at stake here.
The only way we can ensure that such action have a valid justification
is by relying on public communications.  That way, everyone knows what
we are doing, and why.  Also, we are much less likely to make mistakes
this way, as people get a chance to correct errors in arguments before
decisions are made.

>	Your vindictive response of further delaying the patch does not
>help GCC or anyone.  You yourself were one of the most outspoken critics
>of the previous GCC development delays and inability to get patches into
>the FSF sources.  How ironic that you do not see yourself falling into the
>same trap and creating the same situation.

My dictionary defines vindicitive as "disposed or inclined to revenge".
I offered to help you get the patch into the FSF sources.  I fail to see
how that constitutes an act of revenge.

I did qualify the form of my assistance, saying that there would be a one
week delay.  This is not intended to be a revengeful act.  It is intended to
be a disciplinary act.  As I see it, you did two things wrong
1) You made a major decision based on secret communications.
2) You made veiled accusations that were easily misinterpreted.
I do not wish to see these things continue, so I choose an action that I
hoped would discourage you from repeating these things in the future, while
at the same time offering help to resolve your problem.

>You yourself were one of the most outspoken critics
>of the previous GCC development delays and inability to get patches into
>the FSF sources.

I was vocal about patch review problems before the EGCS project started.
Things are much better now.  However, no matter how much we improve the
process, there will always be problems.  Not everyone that wants to contribute
to gcc has sufficient technical skill to write good patches.  Not everyone
that wants to contribute will have a personality that meshes with other
gcc developers.  Not every patch contributed is a good idea.  There will
always be patches that slip through the cracks no matter how good the process
is.  Overall, I think we are doing a reasonable job.  People who are
persistent, who do consistently good work, and who show a willingness to
finish what they start, generally are able to get their patches in.  People
who are casual contributors will likely have problems, but that is a tough
problem to solve without endangering the stability of the sources.  At least
we have the GNATS database now, which gives us a place to put patches where
they won't be lost.  I don't know how to improve on that without hiring
someone to do the task, and we don't have money for that.

>How ironic that you do not see yourself falling into the
>same trap and creating the same situation.

I fail to see how I am adding to the problem.  After all, I did offer my help
to review the patch.  I could not offer my help earlier, because I was not
privy your secret communications.


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