Simon Peyton Jones writes:

| meantime the issue can be worked around by reverting | accbc242e555822a2060091af7188ce6e9b0144e.

Alas, not so.

git revert accbc242e555822a2060091af7188ce6e9b0144e warning: Failed to merge submodule utils/haddock (commits don't follow merge-base) error: could not revert accbc242e5... DynFlags: disentangle Outputable hint: after resolving the conflicts, mark the corrected paths hint: with 'git add <paths>' or 'git rm <paths>' hint: and commit the result with 'git commit'

Sigh, yes, this is what I was afraid of. Strangely, Hadrian's validate flavour doesn't appear to be affected by the issue that you reported. Are you using the make build system by any chance? Cheers, - Ben

| Strangely, `./validate --legacy --slow` also appears to work fine for | me | on 55fd1dc55990623dcf3b2e6143e766242315d757. | | Simon, can you describe how you were previously building GHC? ./validate --legacy --fast with this validate.mk (below). But the issue is terribly simple: assertPprPanic is used (by ASSERT), but no longer imported (by many many modules) because they previously got it from Outputable. How can this possibly work? It certainly doesn't for me. Would it be possible to revert the patch that broke this? I'm fully stalled with no workaround. Thanks Simon SRC_HC_OPTS = -O -H64m GhcStage1HcOpts = -DDEBUG GhcStage2HcOpts = -dcore-lint -ticky GhcLibHcOpts = -O -dcore-lint -ticky BUILD_PROF_LIBS = NO SplitSections = NO HADDOCK_DOCS = NO BUILD_SPHINX_HTML = NO BUILD_SPHINX_PDF = NO BUILD_MAN = NO LAX_DEPENDENCIES = YES | -----Original Message----- | From: Ben Gamari | Sent: 18 August 2020 20:04 | To: Simon Peyton Jones ; GHC developers | Subject: RE: Breakage on master | | Ben Gamari writes: | | > Simon Peyton Jones writes: | > | >> | meantime the issue can be worked around by reverting | >> | accbc242e555822a2060091af7188ce6e9b0144e. | >> | >> Alas, not so. | >> | >> git revert accbc242e555822a2060091af7188ce6e9b0144e | >> warning: Failed to merge submodule utils/haddock (commits don't | follow merge-base) | >> error: could not revert accbc242e5... DynFlags: disentangle | Outputable | >> hint: after resolving the conflicts, mark the corrected paths | >> hint: with 'git add <paths>' or 'git rm <paths>' | >> hint: and commit the result with 'git commit' | >> | > Sigh, yes, this is what I was afraid of. Strangely, Hadrian's | validate | > flavour doesn't appear to be affected by the issue that you | reported. | > Are you using the make build system by any chance? | > | Strangely, `./validate --legacy --slow` also appears to work fine for | me | on 55fd1dc55990623dcf3b2e6143e766242315d757. | | Simon, can you describe how you were previously building GHC? | | Cheers, | | - Ben

| > But the issue is terribly simple: assertPprPanic is used (by | ASSERT), but no longer imported (by many many modules) because they | previously got it from Outputable. How can this possibly work? I believe that that the answer to my qn is "just import GHC.Utils.Panic". More specifically, as part of your patch, in future, if you want to use ASSERTs you must * BOTH #include HsVersions.h * AND import GHC.Utils.Panic and that you have added that extra import to the zillion existing modules that have an ASSERT. I had not understood that change. I think all I need do is to add the extra import to my new modules in !3220. It would be helpful to add a prominent comment in HsVersions.h. I still wonder whether HsVersions.h could simply include 'import GHC.Utils.Panic' to save this clutter. But regardless, I think I'm unstuck now. Thanks Simon | -----Original Message----- | From: ghc-devs On Behalf Of Sylvain | Henry | Sent: 19 August 2020 11:53 | To: ghc-devs@haskell.org | Subject: Re: Breakage on master | | I can't reproduce the issue. Is it on a specific branch? | | If it's in a branch with new ASSERTs, you should just have to import | GHC.Utils.Panic to fix the issue or is there something else? | | Sylvain | | On 19/08/2020 10:24, Simon Peyton Jones via ghc-devs wrote: | > | Strangely, `./validate --legacy --slow` also appears to work fine | for | > | me | > | on 55fd1dc55990623dcf3b2e6143e766242315d757. | > | | > | Simon, can you describe how you were previously building GHC? | > | > ./validate --legacy --fast | > | > with this validate.mk (below). | > | > But the issue is terribly simple: assertPprPanic is used (by | ASSERT), but no longer imported (by many many modules) because they | previously got it from Outputable. How can this possibly work? It | certainly doesn't for me. | > | > Would it be possible to revert the patch that broke this? I'm fully | stalled with no workaround. | > | > Thanks | > | > Simon | > | > | > | > SRC_HC_OPTS = -O -H64m | > GhcStage1HcOpts = -DDEBUG | > GhcStage2HcOpts = -dcore-lint -ticky | > GhcLibHcOpts = -O -dcore-lint -ticky | > | > BUILD_PROF_LIBS = NO | > SplitSections = NO | > HADDOCK_DOCS = NO | > BUILD_SPHINX_HTML = NO | > BUILD_SPHINX_PDF = NO | > BUILD_MAN = NO | > | > LAX_DEPENDENCIES = YES | > | > | > | -----Original Message----- | > | From: Ben Gamari | > | Sent: 18 August 2020 20:04 | > | To: Simon Peyton Jones ; GHC developers | | devs@haskell.org> | > | Subject: RE: Breakage on master | > | | > | Ben Gamari writes: | > | | > | > Simon Peyton Jones writes: | > | > | > | >> | meantime the issue can be worked around by reverting | > | >> | accbc242e555822a2060091af7188ce6e9b0144e. | > | >> | > | >> Alas, not so. | > | >> | > | >> git revert accbc242e555822a2060091af7188ce6e9b0144e | > | >> warning: Failed to merge submodule utils/haddock (commits | don't | > | follow merge-base) | > | >> error: could not revert accbc242e5... DynFlags: disentangle | > | Outputable | > | >> hint: after resolving the conflicts, mark the corrected paths | > | >> hint: with 'git add <paths>' or 'git rm <paths>' | > | >> hint: and commit the result with 'git commit' | > | >> | > | > Sigh, yes, this is what I was afraid of. Strangely, Hadrian's | > | validate | > | > flavour doesn't appear to be affected by the issue that you | > | reported. | > | > Are you using the make build system by any chance? | > | > | > | Strangely, `./validate --legacy --slow` also appears to work fine | for | > | me | > | on 55fd1dc55990623dcf3b2e6143e766242315d757. | > | | > | Simon, can you describe how you were previously building GHC? | > | | > | Cheers, | > | | > | - Ben | > _______________________________________________ | > ghc-devs mailing list | > ghc-devs@haskell.org | > | https://nam06.safelinks.protection.outlook.com/?url=http%3A%2F%2Fmail. | haskell.org%2Fcgi-bin%2Fmailman%2Flistinfo%2Fghc- | devs&data=02%7C01%7Csimonpj%40microsoft.com%7Cee9f099b376042e0fb6c | 08d8442e2b39%7C72f988bf86f141af91ab2d7cd011db47%7C1%7C0%7C637334312373 | 953419&sdata=6XaNrQIfRNqlSnXPAmoxXYTYQorYszHmyJc1kThIU2o%3D&re | served=0 | _______________________________________________ | ghc-devs mailing list | ghc-devs@haskell.org | https://nam06.safelinks.protection.outlook.com/?url=http%3A%2F%2Fmail. | haskell.org%2Fcgi-bin%2Fmailman%2Flistinfo%2Fghc- | devs&data=02%7C01%7Csimonpj%40microsoft.com%7Cee9f099b376042e0fb6c | 08d8442e2b39%7C72f988bf86f141af91ab2d7cd011db47%7C1%7C0%7C637334312373 | 963414&sdata=m2rdARx2tpZFv5I2WiEdYE5d37LeXTMIIPg51Zfee9M%3D&re | served=0