Re: LaMont Jones, WTH are you doing?
On Wed, Feb 06, 2013 at 06:11:03PM +0000, Thorsten Glaser wrote:
> please compare
> http://lists.debian.org/debian-devel-changes/2013/02/msg00141.html
> with, for example,
> http://lists.debian.org/debian-devel-changes/2013/02/msg00140.html
> and tell me whether you’re seeing what I am seeing:
> • As, apparently (he’s done it for bind9 and util-linux too) normal
> for LaMont, he’s
> ‣ uploading (hand-built¹) binaries for two architectures
> ‣ manipulating the .changes file (the order of files is the hint
> here: regular Debian uploads have .dsc and .diff.gz/.debian.tar.gz
> first, his don’t)
mergechanges is responsible for the differences you're seeing:
dpkg-source is run (yes, on an ubuntu system), and then binaries are built
on a system that is running sid, both amd64 and i386 binaries, since at
least one of those buildds has bitten me with bad binaries in the past.
The results are then merged, signed, and uploaded.
> • In addition to that, this time,
> ‣ it’s got an Original-Maintainer field
> ‣ “Maintainer: Ubuntu Developers <ubuntu-devel-discuss@lists.ubuntu.com>”
> … excuse me, what?
That's me failing at merging a change from ubuntu, which will be fixed this
evening in 2.9.6-2. At the moment, fixing the maintainer is the only change
for -2.
> ① hand-built packages are a good guess since #650534 showed that
> the package that was actually uploaded didn’t build on *any*
> platforms
> … though, this time, it looks more like the upload was built on
> an *buntu system and then uploaded to Debian, including binaries?
I think that was the point where I decided that I needed to keep a debian
box around.
> I’ve been seeing this on d-d-changes for a while now. Please explain,
> otherwise I’m going to call for an binNMU (on every binary you uploaded)
> before wheezy is released, since we obviously cannot trust these bina‐
> ries and they’re not exactly leaf packages with < 100 users…
The packages you're seeing are built on a sid amd64 box, the i386 binaries
built in a sid schroot. I suppose I could just trust the i386 buildds to
do the right thing and not screw up, and just upload amd64 binaries to go
with the source.
lamont
Reply to: