Fink:Major New Feature Plans:XMLish Source and Patch

From the Fink Wiki
Jump to: navigation, search


All the *Source*-related fields are a pain to parse, can lead to a messy .info. We could cluster all fields for a given source "item". Easy to add new features (dmalloc wants stronger checksums) without making even more of a top-level fieldname mess.

SourceFile: <<
  Source: http://whatever
  Rename: %n-%v.tar.gz
  MD5: whatever
<<
SourceFile2: <<
  Source: http://whatever
  Rename: %n-%v.tar.gz
  MD5: whatever
<<

Easy to extend to patchfiles (for example, MD5 for patchfiles is a longstanding request).

PatchFile: <<
  Source: foo.patch
  MD5: whatever
<<
PatchFile2: <<
  Source: foo-ssl.patch
  MD5: whatever
<<

The full path to each file would be available as a %-exp and %a would be abolished. That way only files that have passed the MD5 check would be usable.

Extensions

  • vasi:
It would be nice to have a patch folder per package rather than one monster patch.
  • dmacks:
This is already possible (subdirs in %a), and I expect that way would continue to work correctly with the new syntax. One can already even cluster [.info and its .patch] in its own %n subdir.
Make the application of the patches part of the default Patch field instead of an automatic action triggered by the mere presence of the patchfile field (whatever it's called).
  • vasi&dmacks: (on #fink?)
Have a "Substitutions" field in the PatchFile* that lists regexps to apply to the patchfile before applying it to the source tree. That way we can devise a clean solution to the pipefail that plagues the standard "sed ... | patch" approach.
  • We should probably deprecate implicit source around the time we do this. That feature has not been used in 10.4T ever, or 10.3 or 10.2-gcc3.3 in at least a year. This is an annoying special case to support IMO...it means we need hacks like Source:none and Type:nosource instead of just treating "no source listed means there is no Source".