D2596: state: raise ProgrammingError if an invalid key is being accessed

[pulkit (Pulkit Goyal)]

pulkit added a comment.


  In https://phab.mercurial-scm.org/D2596#42818, @yuja wrote:
  
  > In https://phab.mercurial-scm.org/D2596#42678, @pulkit wrote:
  >
  > > In https://phab.mercurial-scm.org/D2596#42672, @durin42 wrote:
  > >
  > > > Hmm, really? Do any states have optional entries that would make this awkward?
  > >
  > >
  > > Nope, but I wanted to be safe here raising a ProgrammingError instead of KeyError.
  >
  >
  > I don't follow, but what if the state file was written by old hg client where some new keys
  >  were missing, and read by new client? Perhaps we'll have to support that scenario.
  
  
  I am going to drop this patch for now so in cases which Yuya mentioned, we can catch the KeyError to realize that key is not available.

REPOSITORY
  rHG Mercurial

REVISION DETAIL
  https://phab.mercurial-scm.org/D2596

To: pulkit, #hg-reviewers
Cc: yuja, durin42, mercurial-devel