[GHC] #8426: one-shot compilation + TH doesn't see instances that is seen in batch mode

[GHC]

#8426: one-shot compilation + TH doesn't see instances that is seen in batch mode
-------------------------------------+-------------------------------------
        Reporter:  errge             |                   Owner:
            Type:  bug               |                  Status:  infoneeded
        Priority:  normal            |               Milestone:  7.12.1
       Component:  Template Haskell  |                 Version:  7.7
      Resolution:                    |                Keywords:
Operating System:  Unknown/Multiple  |            Architecture:
 Type of failure:  Other             |  Unknown/Multiple
      Blocked By:                    |               Test Case:
 Related Tickets:  #7867             |                Blocking:
                                     |  Differential Revisions:
-------------------------------------+-------------------------------------

Comment (by errge):

 Replying to [comment:10 ezyang]:
 > The patches are pretty short; I think the real reason this ticket is
 stalled is because simonpj never signed off on errge's suggested
 implementation strategy; and in the meantime, it seems this bug has been
 routed around. Perhaps we should just apply simonpj's original suggestion
 (remove `[InstanceDec]` from `ClassI` and `FamilyI`) and call this bug
 fixed.

 No, I would like to strongly ask you not to do that.  As I mentioned
 before in the bug report, I depend on the functionality to look up
 instance declarations for classes.  If this info were to be removed, then
 there would be no way for me to continue to support the HFlags library
 after 7.12.

 As can be seen in the history of the ticket in comment 5 I have a patch
 which keeps the functionality in a very careful way (doesn't cause any
 accidental slowness for people who don't want to use it).  It's also easy
 to understand from the user's point of view: looking up instances is
 expensive so you have to call it explicitly if you ever want it.

 I'm prepared to rebase that patch and fix up the test suite, if I get some
 agreement from a maintainer with actual commit rights that the idea is
 acceptable.  Of course, this is not a blank cheque, I would still have to
 provide quality code, but not willing to work a lot again on this just so
 that you guys drop this without ever replying or discussing with me for 14
 months.  I remember prioritizing this work over my other commitments
 exactly so it doesn't stall and keeps moving, but no one cared.  Sorry for
 the harsh words, I can of course totally understand that for everyone else
 the contribution time is already thin and very valuable; this is why what
 I'm looking for is an agreement on goals and designs before anyone (you or
 me) has to invest serious time and effort on either side.

 I'm also open to other implementation ideas, but I seriously depend on
 this feature so would like to ask you guys to keep it on in some form.

 Thanks and sorry for the long comment!

--
Ticket URL: <http://ghc.haskell.org/trac/ghc/ticket/8426#comment:11>
GHC <http://www.haskell.org/ghc/>
The Glasgow Haskell Compiler