Skip to content

require package name permissions to use a dist name #80

wants to merge 27 commits into from

4 participants

rjbs commented Mar 13, 2014

This implements the long-discussed rule:

  • To upload a distribution named "Xyz" you must have permission on package "Xyz."
  • …and it's okay if you only get that permission by the upload under consideration.

The grandfather list needs to be fleshed out. We can do that programatically.

rjbs commented Mar 14, 2014
  • 806 dists can be "fixed" by having a row inserted to assign permissions, though there is no upload
  • 5 more are in a similar state, but do not have dist names that form valid package names; these can probably be inserted as permissions anyway; examples: Net--RabbitMQ, Text-Tabs+Wrap
  • 54 dists can't be obviously made good, and should be left alone for manual correction as needed

Should we consider a more aggressive conversion from distname to distpkg? E.g.


It might be nice to preserve the notion that the perms table is indexed on package names and not arbitrary strings.

andk commented Mar 15, 2014

d0476f4 needs to be reminded that we want to inform the user about what we did. Something in the spirit of 406a89d. All kinds of fails within pause processing need an error category and must find their way to the affected users.

rjbs commented Mar 15, 2014

new permissions to add at

rjbs added some commits Apr 14, 2013
@rjbs rjbs what pkg governs permissions for a dist? 3578aa9
@rjbs rjbs simplify examine_dist, break into subroutines aae3638
@rjbs rjbs stop indexing single-pm-file dists c7b3ec2
@rjbs rjbs refactor more of examine_pms to simplify skimming c1aa224
@rjbs rjbs refactor The Ultimate Sanity Check c2579ed
@rjbs rjbs dbh is not used in parse_version, do not discon 3b57c95
@rjbs rjbs set up transactiona round examine_pms so we can rollback c33b63d
@rjbs rjbs implement the "got permissions?" check for distname perms cd19c50
@rjbs rjbs tests to show that distname permcheck happens (incomplete) 84d6cef
@rjbs rjbs test for error email body looking right b36cdb7
@rjbs rjbs start up the grandfather list... it will get big f883c52
@rjbs rjbs update the refactoring of indexing for new "skip perl-like" code 46bab24
@rjbs rjbs improve diagnostics from PAUSE email tests 5da6d40
@rjbs rjbs correct some test output from index tests
This seems to have been left in a bad state from my last
hunk of work on it! :(
@rjbs rjbs make all mldistwatch tests pass...
...even though I think some of them are wrong-headed atm.
@rjbs rjbs test for a bit more specific text in case-sens error mail 5ccda42
@rjbs rjbs implement file_not_updated_ok 017f4d2
@rjbs rjbs finally implement file_not_updated_ok! cf59816
@rjbs rjbs be more aggressive in converting distname to distpkg
This fixes Net--RabbitMQ and Text-Tabs+Wrap
@rjbs rjbs notice .pm.bz2 files, too! 8cc330a
@rjbs rjbs cope with "my name is 0"
@rjbs rjbs send a skip report for .pm.gz files 2def0ba
@rjbs rjbs avoid newlines in dumped oneline output 79f72c6
@rjbs rjbs test for skip report on .pm.gz file 10598cb
@rjbs rjbs use the logging mechanism, not warn 2b06af7
@rjbs rjbs correct an incorrect expectation 61eefd5
@rjbs rjbs avoid some uninitialized-category warnings 767a03e
rjbs commented Mar 16, 2014

This has been merged, after some problems related to DBH disconnection management. We did use @dagolden's suggestion on conversion, which got us just a few more dists not needing special casing.

@rjbs rjbs closed this Mar 16, 2014
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Something went wrong with that request. Please try again.