Use substTyUnchecked instead of substTy (#11371)
AbandonedPublic

Authored by thomie on Feb 25 2016, 9:57 AM.

Details

Reviewers
niteria
goldfire
austin
bgamari
Trac Issues
#11648
Summary

This fixes TEST='MoreEx T11224b MonoidsTF T11480b T11523' when GHC is
compiled with -DDEBUG.

These tests were broken by commit b5292557dcf2e3844b4837172230575d40a8917e (Trac #11581).

Diff Detail

Repository
rGHC Glasgow Haskell Compiler
Branch
devel2A
Lint
Lint WarningsExcuse: don't bother me with this please
SeverityLocationCodeMessage
Warningcompiler/typecheck/TcPatSyn.hs:255TXT3Line Too Long
Unit
No Unit Test Coverage
Build Status
Buildable 8627
Build 10672: arc lint + arc unit
thomie updated this revision to Diff 6767.Feb 25 2016, 9:57 AM
thomie retitled this revision from to Use substTyUnchecked instead of substTy (#11371).
thomie updated this object.
thomie edited the test plan for this revision. (Show Details)
thomie added a reviewer: niteria.
thomie added a comment.EditedFeb 25 2016, 10:01 AM

@niteria: I don't know what I'm doing, but this is in response to https://mail.haskell.org/pipermail/ghc-devs/2016-February/011455.html. See https://s3.amazonaws.com/archive.travis-ci.org/jobs/110115377/log.txt for the Travis build log.

I see three solutions:

  • push this patch
  • revert b5292557dcf2e3844b4837172230575d40a8917e. (edit: not a008eadfaa4816be349b4fefde9b9b9edc1ca359, as I said earlier)
  • mark the tests as expect_broken
  • fix it properly (but I don't know how)

Which do you prefer?

niteria edited edge metadata.Feb 25 2016, 10:21 AM
niteria added a subscriber: simonpj.

Are you sure that it bisects to a008eadfaa4816be349b4fefde9b9b9edc1ca359? This place looks only vaguely related. If it really broke because of a008eadfaa4816be349b4fefde9b9b9edc1ca359, then my guess would be that a008eadfaa4816be349b4fefde9b9b9edc1ca359 is an incomplete fix.

The place that I marked is supposed to be already fixed, if we push this then I'm not sure how to fix it.

Can @simonpj take a look?

compiler/typecheck/TcHsType.hs
961

This might be a pre-existing problem.

compiler/typecheck/TcPatSyn.hs
255–260

This is a clear regression as indicated by the comment below.

Ah, b5292557dcf2e3844b4837172230575d40a8917e is the offending commit .

thomie updated this object.Feb 25 2016, 11:11 AM
thomie edited edge metadata.
thomie abandoned this revision.Feb 26 2016, 4:53 AM

Alright, I opened Trac #11648 for this, and I'll mark the tests as expect_broken.

simonpj updated the Trac tickets for this revision.Feb 29 2016, 8:31 AM

See Trac Trac #11648