OpenPKG CVS Repository
  http://cvs.openpkg.org/
  ____________________________________________________________________________

  Server: cvs.openpkg.org                  Name:   Thomas Lotterer
  Root:   /e/openpkg/cvs                   Email:  [EMAIL PROTECTED]
  Module: openpkg-re                       Date:   03-Feb-2004 10:15:10
  Branch: HEAD                             Handle: 2004020309151000

  Modified files:
    openpkg-re              todo.txt

  Log:
    shift priorities

  Summary:
    Revision    Changes     Path
    1.170       +46 -5      openpkg-re/todo.txt
  ____________________________________________________________________________

  patch -p0 <<'@@ .'
  Index: openpkg-re/todo.txt
  ============================================================================
  $ cvs diff -u -r1.169 -r1.170 todo.txt
  --- openpkg-re/todo.txt       2 Feb 2004 09:44:58 -0000       1.169
  +++ openpkg-re/todo.txt       3 Feb 2004 09:15:10 -0000       1.170
  @@ -40,18 +40,17 @@
       o fix bootstrap %pre problem [rse]
         buildfarm reports "install: %pre scriptlet failed (2), skipping openpkg ..."
         (currently does no longer occur?)
  -    o openpkg.boot does not find rpmtool through macro while bootstrapping [rse]
       o companion release GNU shtool 2.0.0 [rse] thl: sh.* docs done
       o check rpm 1.3/2.0 package cross signing and verification [thl]
       o make sure openpkg-2.0.0-2.0.0 provides: openpkg-20040113-20040113
       o decide whether *-2.0.0.(src.)rpm should require: openpkg-2.0.0-2.0.0 or not 
and why (not)
       o info.dir (Debian v3.1), /override/bin, wrappers for cp|mv|rm|chmod|chown|chgrp
       o fix all CORE and BASE packages
  -      
http://www.openpkg.org/status-log.cgi/aide-0.9-20030724.log.dv16.ix86-solaris10-re [ms]
  -      checking for unsigned short... yes
  -      checking size of unsigned short... configure: error: cannot compute sizeof 
(unsigned short)
   
     SHOULD HAVE:
  +    o <compat> not <loc>
  +    o use UUID for COOKIE header
  +    o openpkg.boot does not find rpmtool through macro while bootstrapping [rse]
       o fix all PLUS packages that existed in OpenPKG 1.3; downgrade to EVAL/JUNK as 
a last resort [ms]
       o Version: %{release} deploy: problems! -> openpkg build! Release: %{version} 
XX, Version: %(rpm -q --qf '%release' *.spec)) [thl]
       o rse: simplify options (for tools, etc):
  @@ -72,7 +71,7 @@
       - thl: basic set of meta-xxx packages
       o fix all PLUS packages or downgrade to EVAL/JUNK [ms]
       o fix EVAL/JUNK packages and elevate to PLUS [ms]
  -    - rse: OpenPKG instance management autit log: added/removed/upgraded packages
  +    - rse: OpenPKG instance management audit log: added/removed/upgraded packages
       - rse: RPM extension: fetch/curl replacement (OSSP fetch)
       - thl: news.txt replaced by news.cgi which uses pre-scanned information
       - rse: architecture/platform independent packages (BuildArch: noarch, aber 
BuildOS: any existing nicht)
  @@ -84,6 +83,48 @@
       - openpkg-dev creates /// in nosrc pathnames
       - rc should check whether it needs *and* can use su to avoid nightly error 
mails from
         cron in instances that were installed non-root (s_usr != root)
  +    - boostrap should remember things from install time, i.e. did it
  +      create users? This is needed to decide whether or not the users
  +      must be erased on uninstall.
  +
  +  REJECTED SCOPE CREEP (but needs to be discussed post release)
  +
  +    o more accurate version.release requirements (Christoph supported by Thomas)
  +    
  +      currently all release packages require other packages unversioned,
  +      i.e. the relationship of application A requiing library L is
  +      formulated as
  +
  +        A requires L
  +
  +      This is good as long as the user stays within a release. However,
  +      when the user starts to mix in CURRENT or is in the progress of
  +      doing an UPGRADE it would be favourable when the requirement is
  +      formulated:
  +      
  +        A requires L >= V.V.V-R.R.R
  +        A requires L >= *-R.R.R
  +
  +      This would ensure L is upgraded before A, something we cannot
  +      assure today. The first form doesn't work for mixing in CURRENT
  +      beyond the first vendor upgrade. It also doesn't work for doing
  +      an UPGRADE assumed the later release also contains a later vendor
  +      version. The second form doesn't work because RPM doesn't support
  +      it.
  +
  +      A possible workaround was identified to omit the vendor version:
  +
  +        A requires L >= L.release.R.R.R and
  +        L-V.V.V-R.R.R provides L.release.R.R.R
  +
  +      This means that every application A and every library L needs
  +      to have this information added. This is a maintenance nightmare
  +      and, if done manually, considered being a very error prone task.
  +      Ralf tried to automate it using macros but failed because the
  +      macro did only get information from the last requirement entry
  +      not all entries and conditional entries seem to be a even harder
  +      challenge. So we stick to what we had in the past and this is
  +      deferred due to time constraints.
   
     OpenPKG 2.1 Development
     -----------------------
  @@ .
______________________________________________________________________
The OpenPKG Project                                    www.openpkg.org
CVS Repository Commit List                     [EMAIL PROTECTED]

Reply via email to