#1875 document noarch rpmdiff behaviour
Merged 2 years ago by tkopecek. Opened 2 years ago by tkopecek.
tkopecek/koji issue996  into  master

file modified
+18
@@ -123,6 +123,24 @@ 

  Note: is a single entity, so denote multiple arches within quotes (e.g.

  'i386 i586 i686').

  

+ How noarch sub-packages are built

+ ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~

+ 

+ It is a technical detail, but can make sense for someone debugging

+ noarch issues. If there are multiple architectures defined, multiple

+ `buildArch` tasks are spawned. Each of them will produce arch and also

+ noarch packages. While arch packages are unique compared to other

+ subtasks, noarch packages should be the same. It shouldn't matter on

+ which arch you're building it. If there is a difference, it is a bug.

+ Koji internally checks some data from these noarch packages and

+ chooses randomly one, which will appear in final builds. If

+ differences are found, build will fail. Koji is using simple embedded

+ rpmdiff variant, which checks for each file included in rpm: mode,

+ device, vflags, user, group, digest. Furthermore, Name, Summary,

+ Description, Group, License, URL, PreIn, PostIn, PreUn, PostUn,

+ Provides, Requires, Conflicts and Obsoletes are compared.  In case of

+ failing this test, `BuildError` is raised.

+ 

  Manually submitting a task

  ~~~~~~~~~~~~~~~~~~~~~~~~~~