- User Since
- Jun 6 2014, 2:22 AM (224 w, 3 d)
Fri, Sep 21
- Fix closure size
Thu, Sep 20
I did some refactoring but haven't slow validated yet. Marking as "changes planned" to avoid merging in the meantime.
- Comments and fixes
- Revert AP_STACK_NOUPD change
- Implement a variant of OVERWRITING_CLOSURE that takes a size arg
- Revert AP_STACK_NOUPD change
Thanks! LGTM but added an inline comment.
Wed, Sep 19
Tue, Sep 18
Sorry for the noise -- I tried again and this seems to fix the segfaults. I guess I used an older binary of the app before.
Just a note: the reproducer still segfaults with this patch.
Mon, Sep 17
Sun, Sep 16
Indeed this introduces 14 failures. Marking as "requested changes" to avoid merging.
Thanks. This looks good to me. There are some test suite failures so we should check those are existing errors or introduced with this patch.
Fri, Sep 14
Wed, Sep 12
I updated the ticket -- there are still a lot of segfaults and assertion errors.
This seems to fix the segfault when I apply this with the STM patch, but the reproducer now fails with another error (see the ticket).
Tue, Sep 11
@simonmar This seems to work fine for stg_GCD_CAF but I don't understand Cmm calling conventions too well so I don't know if this change makes sense. In particular I assumed that the object that we entered is always in R1. Could you take a look?
Mon, Sep 10
Sun, Sep 9
@simonmar anything else?
Sat, Sep 8
Introducing the new DynFlag isInteractive is a first step in this direction
So unless I'm wildly misunderstanding here
Fri, Sep 7
Why not simply change the error message to
@simonmar I deleted the ghci test as ghci is also using threaded runtime which causes this test to hang.
One of the reasons why I wanted reviews for this patch is because it changes visible arities of some definitions (e.g. pooledReallocArray previously had arity 0, now it has arity 3) and that may effect inlining decisions. I don't think it'll cause any problems downstream but I'm not sure. Any opinions @dfeuer @RyanGlScott @bgamari ?
Thu, Sep 6
- Remove comment
Ah! I see now that -po does not imply -p. That's not too clear in the man page but it's clear in the user manual. Sorry for the noise.
Thanks for doing this! I haven't reviewed this in detail but you can't make constructor as it's currently a valid identifier and making it a keyword breaks programs that use it as an identifier. In fact, I can't even boot GHC with this patch as some of the ghc libraries use constructor as an identifier.
Wed, Sep 5
Tue, Sep 4
Mon, Sep 3
Fri, Aug 31
- Add prototype
Thu, Aug 30
Ah so you wanted to update languagePragParseError too? Why? To be clear, your wording is fine (but note that I'm also not a native speaker ;-), my concern is about consistency and not the choice of words of tense.
Thanks @RolandSenn! Added a minor comment about the error message.
Wed, Aug 29
Aug 22 2018
I reverted this commit with ab55b4ddb7.
@bgamari this should be reverted as it breaks another program (not caught by the test suite -- see my comments above). Sorry I should have marked this as "changes planned".
Aug 20 2018
Found it, for reference: https://github.com/ghc-proposals/ghc-proposals/blob/master/proposals/0021-unlifted-array.rst
Once we have UnliftedArray#
Aug 17 2018
Aug 14 2018