Ideas

Shortcuts

• rpm: package
• RS: Required Services – services, required by pkg (rpm -q --requires rpm | grep -v ^/ | gawk '{print $1}')
• RF: Required Files – files, required by pkg (rpm -q --requires rpm | grep ^/)
• PS: Provided Service – services, provided by pkg (rpm -q --provides rpm | gawk '{print $1}')
• PF: Provided Files – files, provided by pkg (rpm -q --provides rpm | gawk '{print $1}')
• 1, ?, +, *: multiplycity (1, 0 or 1, 1+, 0+) – qty of the service/file in 1 rpm or all rpms

?

• rpm: can B several rpms w/ the same name, but different versions.
• RS: 0|1(?)/rpm; can require own PS (loopback)
• RF: */rpm; loopback
• PS: 0|1(?)/rpm; */all (can be provided by several rpms)
• PF: 0|1(?)/rpm; */all

??

• result includes only need services and files

???

• Input:
  • get all rpms,
  • their RS, RF (splitting --requires by heading /)
  • PS, PF 
  • all – in strings
  • make sort|uniq
  • duplicate rpms – w/ version
• Resort:
  • resort all names
  • make Name>ID hashes
  • delete unwaned
• Output:
  • char *rpm[] – array of all rpm names; == ID>Name
  • char *svc[] – array of all required services names
  • char *file[] – array of all required files names
  • int g2r[rpms][] – array of rpm requires IDs 
  • int g2p[rpms][] – array of rpm provides IDs 
  • int g2f[rpms][] – array of rpm files IDs 
  • int r2g[rpms][] – array of required by rpms IDs 
  • int p2g[rpms][] – array of provided by rpms IDs 
  • int f2p[rpms][2?] – array of files to rpms IDs

Tests

PIII-1000, 1091 pkgs

Thoughts

7324

Notes:

• rpmorphan.cc
  • check correct list
  • -ver-rel – for dupes only
  • check loopbacks
• rpmtop:
  • split on 3 utils:
    • rpm-all (типа rpm -qa)
    • rpm-orphqned
    • rpm-explorer (w/ gui/cli, “what if”, protected etc)
  • and then – merge into 1 utility (but explorer – w/ reading rpm db each of recuring step).

Algo:

• Stage #1: read all pkgs; for each:
  • name – check for dup (on dup and global pkgs) and correct – own and previous; add to “dup” set.
  • provides:
    • local set of provides + vector of provides (=> global pkgs)
    • local set of files + vector of files (=> global pkgs)
  • requires
    • split on service/file
    • if not in local provides/files and not in global required – ad to global required; but not save in global pkgs
• Stage #2:
  • read pkgs; for each:
    • find provides in global required; if found – mark this pkg as “wanted” and break
    • the same – in files
    • if not break – add this pkg to “orphaned” (or just out to console)
• Stage #3: out to console

Optimize:

• отлов required при чтении provided – пометили – проверили required – и сносим
• все строки – в отдельный массив
• вместо set/map обходиться массивами + map (name/ID)
• 2 прохода по rpm (или запоминать сцылки)? – но это по 1.5 сек за раз. Зато память экономится.