New File Handling

Terminology:

• MF - Metadata file - File listed in repomd.xml.
• OF - Other file - File inside repodata/ directory but not listed in repomd.

Current (old) behavior

Regardless if used with or without --update createrepo(_c) output repodata/ looks like:

• All OFs stay there with unmodified content.
• Some MFs are regenerated. According to command line parameters at least the primary.xml, filelists.xml, other.xml, but other MFs like groupfile (--groupfile) or databases (--database) could be regenerated as well.
• Other MFs that were listed in repomd.xml are kept in the repodata/ directory but not referenced by repomd.xml. (See: https://bugzilla.redhat.com/show_bug.cgi?id=950301)
• The only files that could be removed are primary, filelists, other and its databases.
• This behavior could leads to the situation where old and outdated files like groupfiles or updateinfos are accumulating in the repodata/ directory over the time.

New behavior

Regardless if used with or without --update, createrepo(_c) output repodata/ should look like:

• All OFs stay there with unmodified content, but each file is a copy so the file's metadata like mtime, or number of hardlinks may be differ.
• All MFs are regenerated or removed. (All files that were referenced by the previous repomd.xml but in the current version they aren't present, are removed from the repodata/ directory)
• All MFs are regular files, not symlinks.
• All MFs have a number of hardlink references equal to 1.

Benefits of new behavior

• Clearly and simple defined
• Users know what they could expect
• More clear and logical than the old behavior