Re: CEDET Merge

[Edward John Steere <edward.steere@gmail.com>]

> Thanks for your detailed response.
>
>> The problem which I face, without the help of those more experienced
>> than I, is that I don't have enough context on how to resolve some of
>> the problems which arise as part of doing a partial merge of CEDET into
>> core.  I'm not an expert wrt to the CEDET project -- I'm a user and an
>> occasional hacker who has an understanding of the project and wants to
>> give back, help it grow and enjoy the benefits of this growth and
>> improvement to Emacs.
>
> Can you say which parts of the merge caused difficulty?  I'd like to
> help move the process along, if it's possible (though I'm also not a
> frequent Elisp hacker).

That's great.  I'd be interested in collaborating with you.

I think that the biggest problem which I faced was in untangling the
dependencies which had been created between code in files which exist in
both core and upstream and some of the newer files which only exist in
upstream.  Some of these dependencies don't seem to be correct.  For
example; there is now an indirect link between the parser and java
support and not in the downwards direction but from the parser up to
java support (!?)

Others dependencies appear to be valid but exist between code which
should be merged and which shouldn't.  This made my approach of trying
to preserve history very challenging because I not only had to ensure
that the roots of patches from CEDET matched the correct destination in
their new home, but I also had to resolve conflicts where they diverged
and simultaneously remove dependencies where they were built up between
those parts which were destined for core and those which were not.

Dependency issues caused problems when trying to bootstrap Emacs on my
branch which had the whole of CEDET merged.  I did receive some help
from Eli at that point, but didn't get very far before David got
involved and started to drive that side of things.

Unfortunately, this vague description is the best I can do without
diving in again.

> Also, do you know which parts of CEDET were going to be offered as
> ELPA packages?  You mentioned COGRE already -- were there any others?

What follows is a (possibly non-exhaustive) list of the parts of CEDET
which I don't think were intended to reach core (gathered by diffing
trees and combining that with what I remember).  This is based on the
premise that if it wasn't in core previously then it shouldn't be merged
-- which is what was agreed upon.  Looking at the list though this does
exclude a lot of good stuff.
  - Everything in contrib (since I don't think that any of it is in
    core) including:
    - eassist
    - ede-gnustep
    - semantic-ectag-scala
    - semnatic-tag-folding
    - wisent-csharp
    - wisent-php
    - wisent-ruby
  - cedet-java and related including:
    - java-tags
    - java
    - jvm-base
  - cedet-graphviz
  - Various ede extensions
    - ant
    - maven
    - compdb
    - android
    - java-root
    - arduino
    - lein2
  - parts of db:
    - db-cscope
    - db-mozrepl
    - db-javap
    - db-mk
  - support for ectags:
    - ectags/util
    - ectags/lang2
    - ectags/parse
    - ectags/lang
    - ectags/db
  - more bovine parsers:
    - clang
    - f90
    - erlang
  - canned configs
  - db-search