[Date Prev][Date Next] [Thread Prev][Thread Next] [Date Index] [Thread Index]

Re: please put fai 2.8.2 into testing/sa



Hi,

On Sunday 08 May 2005 02:56, Steve Langasek wrote:
> While I'm sure all of the changes you've made are justified in their own
> right, we can't justify spending release team time on reviewing changes
> that don't meet the criteria for freeze exceptions.  Please, upload a
> version this package fixing only the RC/documentation/translation issues to
> testing-proposed-updates.

Hmmm... starting something with "i agree... but..." usually doesn't really 
indicate agreement or understanding. But it's true and we do understand your 
point and agree, but we also would like to point out some facts you might not 
have been aware off:

- fai-2.8.2.dif contains only RC-bugfixes. 

- fai-2.8.1 was uploaded on April, 30th and blocked by manual hint by aba 
which is commented with " # upload urgency inflation ..." - but the urgency 
of the 2.8.1 was low. Without that hint, fai 2.8.1 would have made it into 
sarge.

- fai-2.8.1 was heavily tested and reviewed by a group of nine fai-developers 
and heavy-users who gathered together in a fai workshop in the beginning of 
april (the work for 2.8.1 was done on a list, 2.8 was created at the 
workshop.)

- lots of documentation changes (alone 550 lines of the 1100 lines diff from 
2.8 to 2.8.2 belong to typo fixes in the guide) -  including changes 
in /usr/share/doc/examples/, which in the source package reside in 
the /examples directory 

-  a few bugs which were fixed in 2.8.1 were not marked correctly as important 
(including the bugs which are about issues on amd64 which also happen on ppc 
+ $otherarchs) - this means, not only would we have to create a fai-version 
2.8.2_0sarge1 (which is a lot of work just to re-introduce bugs - 2.8.1+2.8.2 
are bug-fix-releases) but we would also have to reopen those bugs, mark them 
RC and fix them again. Examples are #304824, #305861, #307147, #306264, 
#305861 - plus there are some more but without bugnumbers as they were fixed 
when they were found. (Remember, we had a developers meeting.) - Obviously 
acting like this was wrong, and we (should) have learned from that: file bugs 
before you fix them (if the release is close like it was in the beginning of 
april) and raise the severity if applicable. (Don't try to be nice. This has 
been a misunderstanding of howto be nice. Be honest. Much nicer ;)

- most if not all other bugfixes are really absolutly trivial - trivial means 
one change in one line. Not 15 lines semi-trivial stuff ;-) 

- fai has no external dependencies and can/could go in thru unstable.

- below you find the comments to the diff again (slightly modified (made  
clearer) compared to the file diff-comments which was linked in Thomas last 
mail.)


We absolutly agree, that fixing so many bugs (at this time) without proper 
bugnumbers is bad and has put unneccessary burdon on you, the release 
managers.  Sorry for that!

On the other hand, our priority are our users. It really makes no sense (to 
us) to advice them to install 2.8.2_0sarge1 from sarge if they can get 2.8.2 
from the fai homepage with no more features but only less bugs.

But as said in the beginning, we understand your priorities, so if you still 
say "no way, please backport important and above bugfixes and doc changes", 
we'll (unhappily) do this. 

regards,
        Thomas Lange
        Holger Levsen
	Henning Sprang
	Michael Tautschnig
	Juergen Kahnert

So here, again, is the list of changes from 2.8 to 2.8.2, you'll find the 
corresponding diff at 
http://www.informatik.uni-koeln.de/fai/download/diff-2.8-2.8.2
(And yes, changes without a # don't have a # :-(( But those changes are really 
simple at least.)

Lines 1 to 30
    only document changes

Lines 31 to 44
   * make-fai-nfsroot.conf: remove --arch i386 option, since debootstrap now
    uses dpkg --print-installation-architecture instead of another call which
    needed gcc, also a small change in comment
    (this is important for all archs !i386)

Lines 45 to 102
    only the changes of the changelog

Lines 103 to 112
    * control: add grub to suggests

Lines 112 to 134
 * debian/postinst: replace several sed -i calls with perl -pi -e
 since old sed does not support -i, we like to support the installation of the 
fai
 package on a woody system, so you can install a new sarge system from
 an old woody system

Lines  143 to 169
  * de.po, templates.pot: remove fuzzy lines, otherwise the debconf
  translation does not work

Lines 170 to 185
  README.disk_config: documentation changes

Lines 186 to 742
  documentation change: this is the very long change, but only small typos in
  the fai guide are fixed.

Lines 743 to 762
  documentation change: some lines from this script are moved to a
   separate script which is called 50-host-classes

Lines  763 to 772
  documentation change: only fixes a typo

Lines  773 to 786
       This fixes bug 305861. A variable name had a typo. Even this
       but was only classified a normal (since it does not occure when
       using lilo), I think this was more a important bug, since default
       installation will use GRUB as bootloader and without this fix
       these installations can not boot.

Lines  787 to 794
       documentation change:  adds a string to a grep pattern to look for 
       errors. This will not influence the installation itself, but only the
       summary of possible errors in errors.log. This may be called a minor
       enhancement, but it will definitely not break anything.

Lines  795 to 813
      documentation change: Do not call tune2fs when using 2.6 kernel. This
      may hang the system during an installation and is IMHO a bug of the
      kernel. This workaround is now documented in bug 307147. 

Lines  814 to 827
         documentation change: Add two arguments when calling a certain
         script. This will mount the local hard disk partitions read only.     
         Before they were mounted rw which was a bug, but not yet reported.

Lines  828 to 843
       trivial: set FAI_ACTION to noaction if undef (closes: #305283)

Lines  844 to 858
       use mktemp for creating temp file, this closes a serious
       security bug (closes: #307838)

Lines  859 to 873
       documentation change only

Lines  874 to 888
       documentation change only (closes #304980)

Lines  889 to 937
       fai: create dir only during initial installation, which fixes a
       critical bug (closes: #307632), also in 2.8.1 the variable
       $COLOR_FAI_LOGO was added which has the same function as the
       old file /.nocolorlogo before

Lines  938 to 953
       test if the command mkinitrd-cd is available

Lines  954 to 1007
       * fai-mirror: fix wrong path (closes: #304179), remove needless
       test if mkinitrd-cd is available, replace "-qq" with the
       variable $qflag (which is just "-qq"), return if $pkglist is
       empty, not when $addpackages is empty. This minor bug was just
       send to me not filed to the BTS.

Lines  1008 to 1029
       change indentation of text message

Lines 1030 to 1059
     * install_packages: add missing aptitude-r call (closes: #305544), 
    replace -qq with $qopt (which is "-qq") and "/dev/null/ with $devnull, 

Lines 1060 to 1077
      fix typo in variable name (closes: #304824) - bugfix is important as
      it removes /floppy!

Lines 1078 to 1096
      split packages list for i386 and amd64 and remove a packages
      from amd64 which are not available (closes: #306264)
      This is also a needed on others archs.

Lines 1097 to 1110
      do not "mount --bind" when FAI_ROOT=/, closes important bug (closes: 
#307631)

Attachment: pgp6BR78cgQUp.pgp
Description: PGP signature


Reply to: