Re: Debian needs more buildds. It has offers. They aren't being accepted.
Riku Voipio <email@example.com> writes:
> On Fri, Feb 13, 2004 at 10:50:50AM +0100, Ingo Juergensmann wrote:
> > On Fri, Feb 13, 2004 at 03:32:56PM +1000, Anthony Towns wrote:
> > > > However working around a lagged and unresponsive port is a lot
> > > > harder.
> > > None of the ports are significantly lagged at the moment.
> > Other people do think different about that.
> Yes, but Release Mainteners opinion has a lot more weight. The situation
> NOW is a lot better than it was a few days before this flamefest begun.
> > > They shouldn't stir up trouble when there simply
> > > isn't a serious problem...
> Judging from all the evidence that is available publicly ( buildd.d.o,
> packages.qa.d.o, bjorn.haxx.se/debian etc) There was and for lesser
> extend still is a serious problem. Most of > 10d waiting packages were
> either Out of Date or waiting for something that is Out of date.
> (Actually this seems changed slightly since I last checked, perl
> is holding a lot because of A RC bug that doesnt seem that RC,
> and there has appeared a RC bug in the alsa-jack + rdpes chain).
> Since you disagree, I would like to know what I have missed,
> what is the information I am overlooking that would tell me that it
> is obvious that there was no serious problem.
> > > > Just to make myself clear, most buildd maintainers do a spectacular
> > > > job, all my (mostly stupid) questions and requests have been handled
> > > > timely.
> > > ...as you point out.
> > I believe you missed the little word "timely" now or ignore the qt-x11-free
> > problem lately. Otherwise you should clarify your definition of "timely",
> > please.
> Emphasis on the "my" part. I know other people have had more trouble,
> but I dont like spreading secondhand information.
> I didnt request for qt-x11-free on mipsel since when I noticed it being
> stuck, there was already a recent request(s!) on depian-mips list.
> I asked if anyone knew who had an an mipsel machine around. I was
> directed to Goswin, who told me there should be no need to panic
> since qt was already on the top of needs-build list, and should
> thus be compiled soon. I disagreed, since I remebered seeing
> qt week earlier in the pole position too. Goswin told that I
> must have remembered wrong since the need-build list is fifo.
> Since my memory occasionaly leaks, I assumed that he was right.
> Few days later I came back asking howcome nothing has changed, but
> at that point qt was already being built.
To clarify that a bit for those not involved or familiar:
When a source is uploaded and enters the accepted queue wanna-build
picks it up and adds it to a list of packages: "needs-build". The list
is _NOT_ FIFO. Packages get added at different places in the queue
depending on certain factors.
Each buildd asks wanna-build what to build, gets a few packages and
starts building them (the package gets marked "building"). Now some
buildds shouldn't build certain packages (e.g. packages that require
more harddisk space or ram than available) and they give those back to
wanna-build so a different buildd can pick up the package.
So when I saw that qt was stuck I checked with wanna-build and saw
qt-x11-free being the first on the list. So unless somone uploads an
important package it would have been the next to build. So no worries
there (or so I thought).
But the buildd allways returned qt-x11-free because it was in
weak_no_auto_build (means only build when your idle) [as martin has
found out now, not Ryan]. So qt stayed at the top of the needs-build
list forever since there is no other buildd to pick it up and the one
buildd was never idle (due to the backlog).
After seeing that behaviour for some time and no reaction from Ryan a
manual build was called for.
> OTOH, at the same time I asked why the arm-version of qt-x11-free had
> been uploaded but not installed for a few weeks. The person responsible
> for the buildd told me that he would look at it and soon enough told
> me that the changes where just signed and accepted.