Sync head; ADTs for parser errors

[shayne-fletcher]

• Sync to https://gitlab.haskell.org/ghc/ghc.git 70dc2f09a33a4c3f485d8b63e92a21955643a0b7
• Adapt parser error handling on HEAD to https://gitlab.haskell.org/ghc/ghc/-/commit/a5aaceecaa04ce7ea5bade6eb96c0d129109c15a

[hurryabit]

This looks very reasonable. Thank you.

[hurryabit]

/azp run

[azure-pipelines]

Azure Pipelines successfully started running 1 pipeline(s).

[shayne-fletcher]

the upstream msys glitch in the last test run on windows seems to be overcome as of today. if this could be /azp run again please testing should succeed this time. argh, nope remains broke.

[shayne-fletcher]

the issue preventing the windows test from passing appears to be this msys2/MSYS2-packages#2171

[hurryabit]

/azp run

[azure-pipelines]

Azure Pipelines successfully started running 1 pipeline(s).

[shayne-fletcher]

well, that was exciting 8.8.4 linux master failed in that run with

2020-10-07T12:48:15.2052734Z [147 of 275] Compiling GHC.Settings.Constants
2020-10-07T12:48:15.2053179Z 
2020-10-07T12:48:15.2053651Z <no location info>: error:
2020-10-07T12:48:15.2054789Z     ghc: panic! (the 'impossible' happened)
2020-10-07T12:48:15.2055627Z   (GHC version 8.8.3 for x86_64-unknown-linux):
2020-10-07T12:48:15.2058643Z 	Unknown IfaceDecl tag: 56
2020-10-07T12:48:15.2058826Z 
2020-10-07T12:48:15.2059121Z Please report this as a GHC bug:  https://www.haskell.org/ghc/reportabug
2020-10-07T12:48:15.2059377Z 
2020-10-07T12:48:15.2523547Z [228 of 275] Compiling GHCi.BreakArray

never seen that happen before :-/ anyway, the msys ticket reports that the server issue is resolved now. once more time with the /azp run should do it. thank-you.

[shayne-fletcher]

Argh. Now the server is back but the content is missing.

2020-10-08T12:12:40.5991346Z configure: Checking for Windows toolchain tarballs...
2020-10-08T12:12:41.7013533Z #=#=#                                                                         
2020-10-08T12:12:41.8117121Z ##O#- #                                                                       
2020-10-08T12:12:41.8117888Z 
2020-10-08T12:12:41.9390589Z ##                                                                         3.8%
2020-10-08T12:12:42.0658904Z ############                                                              17.0%
2020-10-08T12:12:42.1908468Z ######################                                                    31.0%
2020-10-08T12:12:42.3156198Z ###############################                                           43.7%
2020-10-08T12:12:42.4373488Z #########################################                                 57.7%
2020-10-08T12:12:42.5398881Z ###################################################                       71.2%
2020-10-08T12:12:42.6498857Z #############################################################             85.8%
2020-10-08T12:12:42.6580757Z #####################################################################     96.6%
2020-10-08T12:12:42.6585368Z ######################################################################## 100.0%
2020-10-08T12:12:42.9681662Z #=#=#                                                                         
2020-10-08T12:12:43.0704395Z ##O#- #                                                                       
2020-10-08T12:12:43.1799238Z ##O=#  #                                                                      
2020-10-08T12:12:43.1800738Z #=#=-#  #                                                                     
2020-10-08T12:12:43.1803828Z curl: (22) The requested URL returned error: 404 Not Found
2020-10-08T12:12:43.2236202Z 
2020-10-08T12:12:43.2238255Z ERROR: Download failed.

[shayne-fletcher]

i'm inclined to comment out that build from the pipeline while msys remains in this state. it is after all primarily testing that we can generate a ghc-lib on windows; as it happens, that's not that useful. there are other places where it is tested that ghc-libs build on windows. comments?

[cocreature]

/azp run

[azure-pipelines]

Azure Pipelines successfully started running 1 pipeline(s).

[shayne-fletcher]

i'm inclined to comment out that build from the pipeline while msys remains in this state. it is after all primarily testing that we can generate a ghc-lib on windows; as it happens, that's not that useful. there are other places where it is tested that ghc-libs build on windows. comments?

the last push trials a workaround. we might be in the clear.

[cocreature]

/azp run

[azure-pipelines]

Azure Pipelines successfully started running 1 pipeline(s).