Gents I would like to express my appreciation for the unbelievably quick release of Centos 6.2. Thanks a million! You managed to release 6.2 some 10 days after 6.1. Johnny, you are not that ugly after all :-).
Louis
On Wed, 21 Dec 2011, Louis Lagendijk wrote:
+1
Le 21/12/2011 19:49, Dennis Jacobfeuerborn a écrit :
Hi Dennis,
See wikipedia : http://en.wikipedia.org/wiki/CentOS
I noticed that the release date of 6.2 was updated on this page at least as fast as 6.2 was released ! FWIW, 4.1, 4.2, and 4.3 were released faster, and 4.9 as fast. But I agree, it was a very pleasant surprise to see 6.2 released as fast, after the big delay of 6.0 and 6.1.
Could the developpers explain what change in the process to achieve such a quick release (after a mail a few weeks ago of Johnny explaining how difficult it was now...) ?
And yes, I agree, Johnny and Karanbir are not that ugly after all !
Thanks, Alain
On 12/21/2011 01:27 PM, Alain Péan wrote:
It is still difficult, but all the things that we put into place to build/test 6.1 were still there for 6.2
All the multilib lists, optional channel lists, etc. still worked for 6.2.
And, unlike the older builds, we did not need to build anything outside of the CentOS RPMS, etc.
If you take a look here, SL has documented packages that do not build correctly with only EL6 as the host:
http://ftp.scientificlinux.org/linux/scientific/6rolling/build/koji/targets/
If you look here, there are other things that required special building orders and packages: http://bugs.centos.org/view.php?id=5180
There was also the fact that several packages did not build correctly because of a change in the default environment: https://bugzilla.redhat.com/show_bug.cgi?id=743229
Then there were packages that would build with only a specific version of gcc ... and there are packages that need glibc-static ... and packages with hidden build requiremnts (see the "config_opts['more_buildreqs']" stuff here:
http://ftp.scientificlinux.org/linux/scientific/6rolling/build/koji/configs/...
Those things all needed to be figured out manually, one by one, to get a good build environment.
Now that we have figured those out, with the build environment set, we could just build and test packages and release them.
As I said, the hardest part was getting the outliers and the build system built before ... moving forward it should be much faster now.
On Wed, 2011-12-21 at 14:11 -0600, Johnny Hughes wrote:
That reminds me. I saw this recently on yum update:
Non-fatal POSTUN scriptlet failure in rpm package gtk2 Cleanup : 1:cups-libs-1.4.2-39.el6_1.1.x86_64 146/230 /usr/bin/update-gdk-pixbuf-loaders: line 54: /etc/gtk-2.0/x86_64-unknown-linux-gnu/gdk-pixbuf.loaders: No such file or directory /usr/bin/update-gtk-immodules: line 54: /etc/gtk-2.0/x86_64-unknown-linux-gnu/gtk.immodules: No such file or directory warning: %postun(gtk2-2.18.9-6.el6.x86_64) scriptlet failed, exit status 1
/var/tmp/rpm-tmp.QyhGhF: line 23: /etc/pango/x86_64-unknown-linux-gnu/pango.modules: No such file or directory
The pango warning is an issue with the old package that got removed, but the gtk2 package current on that system (gtk2-2.18.9-6.el6.x86_64 from the CR repo) still has that incorrect directory (/etc/gtk-2.0/x86_64-unknown-linux-gnu/)
And yes, of course, thank you team for all the hard work you put into CentOS!
Regards, Leonard.
-
Alain Péan -
Dennis Jacobfeuerborn -
Johnny Hughes -
Leonard den Ottolander -
Louis Lagendijk -
Paul Heinlein -
Rob Townley -
Shade.GE