[Git][ghc/ghc][wip/andreask/lsp-crash] 43 commits: Refactor the treatment of loopy superclass dicts

Andreas Klebinger (@AndreasK) gitlab at gitlab.haskell.org
Tue Jan 17 14:06:48 UTC 2023



Andreas Klebinger pushed to branch wip/andreask/lsp-crash at Glasgow Haskell Compiler / GHC


Commits:
aed1974e by Richard Eisenberg at 2023-01-11T08:30:42+00:00
Refactor the treatment of loopy superclass dicts

This patch completely re-engineers how we deal with loopy superclass
dictionaries in instance declarations. It fixes #20666 and #19690

The highlights are

* Recognise that the loopy-superclass business should use precisely
  the Paterson conditions.  This is much much nicer.  See
  Note [Recursive superclasses] in GHC.Tc.TyCl.Instance

* With that in mind, define "Paterson-smaller" in
  Note [Paterson conditions] in GHC.Tc.Validity, and the new
  data type `PatersonSize` in GHC.Tc.Utils.TcType, along with
  functions to compute and compare PatsonSizes

* Use the new PatersonSize stuff when solving superclass constraints
  See Note [Solving superclass constraints] in GHC.Tc.TyCl.Instance

* In GHC.Tc.Solver.Monad.lookupInInerts, add a missing call to
  prohibitedSuperClassSolve.  This was the original cause of #20666.

* Treat (TypeError "stuff") as having PatersonSize zero. See
  Note [Paterson size for type family applications] in GHC.Tc.Utils.TcType.

* Treat the head of a Wanted quantified constraint in the same way
  as the superclass of an instance decl; this is what fixes #19690.
  See GHC.Tc.Solver.Canonical Note [Solving a Wanted forall-constraint]
  (Thanks to Matthew Craven for this insight.)

  This entailed refactoring the GivenSc constructor of CtOrigin a bit,
  to say whether it comes from an instance decl or quantified constraint.

* Some refactoring way in which redundant constraints are reported; we
  don't want to complain about the extra, apparently-redundant
  constraints that we must add to an instance decl because of the
  loopy-superclass thing.  I moved some work from GHC.Tc.Errors to
  GHC.Tc.Solver.

* Add a new section to the user manual to describe the loopy
  superclass issue and what rules it follows.

- - - - -
300bcc15 by HaskellMouse at 2023-01-11T13:43:36-05:00
Parse qualified terms in type signatures

This commit allows qualified terms in type
signatures to pass the parser and to be cathced by renamer
with more informative error message. Adds a few tests.

Fixes #21605

- - - - -
964284fc by Simon Peyton Jones at 2023-01-11T13:44:12-05:00
Fix void-arg-adding mechanism for worker/wrapper

As #22725 shows, in worker/wrapper we must add the void argument
/last/, not first.  See GHC.Core.Opt.WorkWrap.Utils
Note [Worker/wrapper needs to add void arg last].

That led me to to study GHC.Core.Opt.SpecConstr
Note [SpecConstr needs to add void args first] which suggests the
opposite!  And indeed I think it's the other way round for SpecConstr
-- or more precisely the void arg must precede the "extra_bndrs".

That led me to some refactoring of GHC.Core.Opt.SpecConstr.calcSpecInfo.

- - - - -
f7ceafc9 by Krzysztof Gogolewski at 2023-01-11T22:36:59-05:00
Add 'docWithStyle' to improve codegen

This new combinator

docWithStyle :: IsOutput doc => doc -> (PprStyle -> SDoc) -> doc

let us remove the need for code to be polymorphic in HDoc
when not used in code style.

Metric Decrease:
    ManyConstructors
    T13035
    T1969

- - - - -
b3be0d18 by Simon Peyton Jones at 2023-01-11T22:37:35-05:00
Fix finaliseArgBoxities for OPAQUE function

We never do worker wrapper for OPAQUE functions, so we must
zap the unboxing info during strictness analysis.

This patch fixes #22502

- - - - -
db11f358 by Ben Gamari at 2023-01-12T07:49:04-05:00
Revert "rts: Drop racy assertion"

The logic here was inverted. Reverting the commit to avoid confusion
when examining the commit history.

This reverts commit b3eacd64fb36724ed6c5d2d24a81211a161abef1.

- - - - -
3242139f by Ben Gamari at 2023-01-12T07:49:04-05:00
rts: Drop racy assertion

0e274c39bf836d5bb846f5fa08649c75f85326ac added an assertion in
`dirty_MUT_VAR` checking that the MUT_VAR being dirtied was clean.
However, this isn't necessarily the case since another thread may have
raced us to dirty the object.

- - - - -
9ffd5d57 by Ben Gamari at 2023-01-12T07:49:41-05:00
configure: Fix escaping of `$tooldir`

In !9547 I introduced `$tooldir` directories into GHC's default link and
compilation flags to ensure that our C toolchain finds its own headers
and libraries before others on the system. However, the patch was subtly
wrong in the escaping of `$tooldir`. Fix this.

Fixes #22561.

- - - - -
905d0b6e by Sebastian Graf at 2023-01-12T15:51:47-05:00
Fix contification with stable unfoldings (#22428)

Many functions now return a `TailUsageDetails` that adorns a `UsageDetails` with
a `JoinArity` that reflects the number of join point binders around the body
for which the `UsageDetails` was computed. `TailUsageDetails` is now returned by
`occAnalLamTail` as well as `occAnalUnfolding` and `occAnalRules`.

I adjusted `Note [Join points and unfoldings/rules]` and
`Note [Adjusting right-hand sides]` to account for the new machinery.
I also wrote a new `Note [Join arity prediction based on joinRhsArity]`
and refer to it when we combine `TailUsageDetails` for a recursive RHS.

I also renamed

  * `occAnalLam` to `occAnalLamTail`
  * `adjustRhsUsage` to `adjustTailUsage`
  * a few other less important functions

and properly documented the that each call of `occAnalLamTail` must pair up with
`adjustTailUsage`.

I removed `Note [Unfoldings and join points]` because it was redundant with
`Note [Occurrences in stable unfoldings]`.

While in town, I refactored `mkLoopBreakerNodes` so that it returns a condensed
`NodeDetails` called `SimpleNodeDetails`.

Fixes #22428.

The refactoring seems to have quite beneficial effect on ghc/alloc performance:

```
     CoOpt_Read(normal) ghc/alloc    784,778,420    768,091,176  -2.1% GOOD
         T12150(optasm) ghc/alloc     77,762,270     75,986,720  -2.3% GOOD
         T12425(optasm) ghc/alloc     85,740,186     84,641,712  -1.3% GOOD
         T13056(optasm) ghc/alloc    306,104,656    299,811,632  -2.1% GOOD
         T13253(normal) ghc/alloc    350,233,952    346,004,008  -1.2%
         T14683(normal) ghc/alloc  2,800,514,792  2,754,651,360  -1.6%
         T15304(normal) ghc/alloc  1,230,883,318  1,215,978,336  -1.2%
         T15630(normal) ghc/alloc    153,379,590    151,796,488  -1.0%
         T16577(normal) ghc/alloc  7,356,797,056  7,244,194,416  -1.5%
         T17516(normal) ghc/alloc  1,718,941,448  1,692,157,288  -1.6%
         T19695(normal) ghc/alloc  1,485,794,632  1,458,022,112  -1.9%
        T21839c(normal) ghc/alloc    437,562,314    431,295,896  -1.4% GOOD
        T21839r(normal) ghc/alloc    446,927,580    440,615,776  -1.4% GOOD

              geo. mean                                          -0.6%
              minimum                                            -2.4%
              maximum                                            -0.0%
```

Metric Decrease:
    CoOpt_Read
    T10421
    T12150
    T12425
    T13056
    T18698a
    T18698b
    T21839c
    T21839r
    T9961

- - - - -
a1491c87 by Andreas Klebinger at 2023-01-12T15:52:23-05:00
Only gc sparks locally when we can ensure marking is done.

When performing GC without work stealing there was no guarantee that
spark pruning was happening after marking of the sparks. This could
cause us to GC live sparks under certain circumstances.

Fixes #22528.

- - - - -
8acfe930 by Cheng Shao at 2023-01-12T15:53:00-05:00
Change MSYSTEM to CLANG64 uniformly

- - - - -
73bc162b by M Farkas-Dyck at 2023-01-12T15:53:42-05:00
Make `GHC.Tc.Errors.Reporter` take `NonEmpty ErrorItem` rather than `[ErrorItem]`, which lets us drop some panics.

Also use the `BasicMismatch` constructor rather than `mkBasicMismatchMsg`, which lets us drop the "-Wno-incomplete-record-updates" flag.

- - - - -
1b812b69 by Oleg Grenrus at 2023-01-12T15:54:21-05:00
Fix #22728: Not all diagnostics in safe check are fatal

Also add tests for the issue and -Winferred-safe-imports in general

- - - - -
c79b2b65 by Matthew Pickering at 2023-01-12T15:54:58-05:00
Don't run hadrian-multi on fast-ci label

Fixes #22667

- - - - -
9a3d6add by Bodigrim at 2023-01-13T00:46:36-05:00
Bump submodule bytestring to 0.11.4.0

Metric Decrease:
    T21839c
    T21839r

- - - - -
df33c13c by Ben Gamari at 2023-01-13T00:47:12-05:00
gitlab-ci: Bump Darwin bootstrap toolchain

This updates the bootstrap compiler on Darwin from 8.10.7 to 9.2.5,
ensuring that we have the fix for #21964.

- - - - -
756a66ec by Ben Gamari at 2023-01-13T00:47:12-05:00
gitlab-ci: Pass -w to cabal update

Due to cabal#8447, cabal-install 3.8.1.0 requires a compiler to run
`cabal update`.
- - - - -
1142f858 by Cheng Shao at 2023-01-13T11:04:00+00:00
Bump hsc2hs submodule

- - - - -
d4686729 by Cheng Shao at 2023-01-13T11:04:00+00:00
Bump process submodule

- - - - -
84ae6573 by Cheng Shao at 2023-01-13T11:06:58+00:00
ci: Bump DOCKER_REV

- - - - -
d53598c5 by Cheng Shao at 2023-01-13T11:06:58+00:00
ci: enable xz parallel compression for x64 jobs

- - - - -
d31fcbca by Cheng Shao at 2023-01-13T11:06:58+00:00
ci: use in-image emsdk for js jobs

- - - - -
93b9bbc1 by Cheng Shao at 2023-01-13T11:47:17+00:00
ci: improve nix-shell for gen_ci.hs and fix some ghc/hlint warnings

- Add a ghc environment including prebuilt dependencies to the
  nix-shell. Get rid of the ad hoc cabal cache and all dependencies
  are now downloaded from the nixos binary cache.
- Make gen_ci.hs a cabal package with HLS integration, to make future
  hacking of gen_ci.hs easier.
- Fix some ghc/hlint warnings after I got HLS to work.
- For the lint-ci-config job, do a shallow clone to save a few minutes
  of unnecessary git checkout time.

- - - - -
8acc56c7 by Cheng Shao at 2023-01-13T11:47:17+00:00
ci: source the toolchain env file in wasm jobs

- - - - -
87194df0 by Cheng Shao at 2023-01-13T11:47:17+00:00
ci: add wasm ci jobs via gen_ci.hs

- There is one regular wasm job run in validate pipelines
- Additionally, int-native/unreg wasm jobs run in nightly/release pipelines

Also, remove the legacy handwritten wasm ci jobs in .gitlab-ci.yml.

- - - - -
b6eb9bcc by Matthew Pickering at 2023-01-13T11:52:16+00:00
wasm ci: Remove wasm release jobs

This removes the wasm release jobs, as we do not yet intend to
distribute these binaries.

- - - - -
496607fd by Simon Peyton Jones at 2023-01-13T16:52:07-05:00
Add a missing checkEscapingKind

Ticket #22743 pointed out that there is a missing check,
for type-inferred bindings, that the inferred type doesn't
have an escaping kind.

The fix is easy.

- - - - -
7a9a1042 by Andreas Klebinger at 2023-01-16T20:48:19-05:00
Separate core inlining logic from `Unfolding` type.

This seems like a good idea either way, but is mostly motivated by a
patch where this avoids a module loop.

- - - - -
33b58f77 by sheaf at 2023-01-16T20:48:57-05:00
Hadrian: generalise &%> to avoid warnings

This patch introduces a more general version of &%> that works
with general traversable shapes, instead of lists. This allows us
to pass along the information that the length of the list of filepaths
passed to the function exactly matches the length of the input list
of filepath patterns, avoiding pattern match warnings.

Fixes #22430

- - - - -
8c7a991c by Andreas Klebinger at 2023-01-16T20:49:34-05:00
Add regression test for #22611.

A case were a function used to fail to specialize, but now does.

- - - - -
6abea760 by Andreas Klebinger at 2023-01-16T20:50:10-05:00
Mark maximumBy/minimumBy as INLINE.

The RHS was too large to inline which often prevented the overhead of the Maybe
from being optimized away. By marking it as INLINE we can eliminate the
overhead of both the maybe and are able to unpack the accumulator when
possible.

Fixes #22609

- - - - -
99d151bb by Matthew Pickering at 2023-01-16T20:50:50-05:00
ci: Bump CACHE_REV so that ghc-9.6 branch and HEAD have different caches

Having the same CACHE_REV on both branches leads to issues where the
darwin toolchain is different on ghc-9.6 and HEAD which leads to long
darwin build times.

In general we should ensure that each branch has a different CACHE_REV.

- - - - -
6a5845fb by Matthew Pickering at 2023-01-16T20:51:25-05:00
ci: Change owner of files in source-tarball job

This fixes errors of the form:

```
fatal: detected dubious ownership in repository at '/builds/ghc/ghc'
To add an exception for this directory, call:
	git config --global --add safe.directory /builds/ghc/ghc
inferred 9.7.20230113
checking for GHC Git commit id... fatal: detected dubious ownership in repository at '/builds/ghc/ghc'
To add an exception for this directory, call:
	git config --global --add safe.directory /builds/ghc/ghc
```

- - - - -
4afb952c by Matthew Pickering at 2023-01-16T20:51:25-05:00
ci: Don't build aarch64-deb10-llvm job on release pipelines

Closes #22721

- - - - -
8039feb9 by Matthew Pickering at 2023-01-16T20:51:25-05:00
ci: Change owner of files in test-bootstrap job

- - - - -
0b358d0c by Matthew Pickering at 2023-01-16T20:51:25-05:00
rel_eng: Add release engineering scripts into ghc tree

It is better to keep these scripts in the tree as they depend on the CI
configuration and so on. By keeping them in tree we can keep them
up-to-date as the CI config changes and also makes it easier to backport
changes to the release script between release branches in future.

The final motivation is that it makes generating GHCUp metadata
possible.

- - - - -
28cb2ed0 by Matthew Pickering at 2023-01-16T20:51:25-05:00
ci: Don't use complicated image or clone in not-interruptible job

This job exists only for the meta-reason of not allowing nightly
pipelines to be cancelled. It was taking two minutes to run as in order
to run "true" we would also clone the whole GHC repo.

- - - - -
eeea59bb by Matthew Pickering at 2023-01-16T20:51:26-05:00
Add scripts to generate ghcup metadata on nightly and release pipelines

1. A python script in .gitlab/rel_eng/mk-ghcup-metadata which generates
   suitable metadata for consumption by GHCUp for the relevant
   pipelines.

  - The script generates the metadata just as the ghcup maintainers
    want, without taking into account platform/library combinations. It
    is updated manually when the mapping changes.

  - The script downloads the bindists which ghcup wants to distribute,
    calculates the hash and generates the yaml in the correct structure.

  - The script is documented in the .gitlab/rel_eng/mk-ghcup-metadata/README.mk file

1a. The script requires us to understand the mapping from platform ->
    job. To choose the preferred bindist for each platform the
    .gitlab/gen_ci.hs script is modified to allow outputting a metadata
    file which answers the question about which job produces the
    bindist which we want to distribute to users for a specific
    platform.

2. Pipelines to run on nightly and release jobs to generate metadata

  - ghcup-metadata-nightly: Generates metadata which points directly to
    artifacts in the nightly job.

  - ghcup-metadata-release: Generates metadata suitable for inclusion
    directly in ghcup by pointing to the downloads folder where the
    bindist will be uploaded to.

2a. Trigger jobs which test the generated metadata in the downstream
    `ghccup-ci` repo. See that repo for documentation about what is
    tested and how but essentially we test in a variety of clean images
    that ghcup can download and install the bindists we say exist in our
    metadata.

- - - - -
97bd4d8c by Bodigrim at 2023-01-16T20:52:04-05:00
Bump submodule parsec to 3.1.16.1

- - - - -
97ac8230 by Alan Zimmerman at 2023-01-16T20:52:39-05:00
EPA: Add annotation for 'type' in DataDecl

Closes #22765

- - - - -
dbbab95d by Ben Gamari at 2023-01-17T06:36:06-05:00
compiler: Small optimisation of assertM

In #22739 @AndreasK noticed that assertM performed the action to compute
the asserted predicate regardless of whether DEBUG is enabled. This is
inconsistent with the other assertion operations and general convention.
Fix this.

Closes #22739.

- - - - -
fc02f3bb by Viktor Dukhovni at 2023-01-17T06:36:47-05:00
Avoid unnecessary printf warnings in EventLog.c

Fixes #22778

- - - - -
4477efff by Andreas Klebinger at 2023-01-17T15:06:30+01:00
ghc-the-library: Retain cafs in both static in dynamic builds.

We use keepCAFsForGHCi.c to force -fkeep-cafs behaviour by using a
__attribute__((constructor)) function.

This broke for static builds where the linker discarded the object file
since it was not reverenced from any exported code. We fix this by
asserting that the flag is enabled using a function in the same module
as the constructor. Which causes the object file to be retained by the
linker, which in turn causes the constructor the be run in static builds.

This changes nothing for dynamic builds using the ghc library. But causes
static to also retain CAFs (as we expect them to).

Fixes #22417.

-------------------------
Metric Decrease:
    T21839r
-------------------------

- - - - -


30 changed files:

- .gitlab-ci.yml
- .gitlab/ci.sh
- .gitlab/darwin/nix/sources.json
- .gitlab/darwin/toolchain.nix
- + .gitlab/gen-ci.cabal
- .gitlab/gen_ci.hs
- + .gitlab/generate_job_metadata
- .gitlab/generate_jobs
- + .gitlab/hie.yaml
- .gitlab/jobs.yaml
- + .gitlab/rel_eng/default.nix
- + .gitlab/rel_eng/fetch-gitlab-artifacts/.gitignore
- + .gitlab/rel_eng/fetch-gitlab-artifacts/README.mkd
- + .gitlab/rel_eng/fetch-gitlab-artifacts/default.nix
- + .gitlab/rel_eng/fetch-gitlab-artifacts/fetch_gitlab.py
- + .gitlab/rel_eng/fetch-gitlab-artifacts/setup.py
- + .gitlab/rel_eng/mk-ghcup-metadata/.gitignore
- + .gitlab/rel_eng/mk-ghcup-metadata/README.mkd
- + .gitlab/rel_eng/mk-ghcup-metadata/default.nix
- + .gitlab/rel_eng/mk-ghcup-metadata/mk_ghcup_metadata.py
- + .gitlab/rel_eng/mk-ghcup-metadata/setup.py
- + .gitlab/rel_eng/nix/sources.json
- + .gitlab/rel_eng/nix/sources.nix
- + .gitlab/rel_eng/upload.sh
- .gitlab/upload_ghc_libs.py → .gitlab/rel_eng/upload_ghc_libs.py
- compiler/GHC.hs
- compiler/GHC/Core/InstEnv.hs
- compiler/GHC/Core/Opt/Arity.hs
- compiler/GHC/Core/Opt/DmdAnal.hs
- compiler/GHC/Core/Opt/LiberateCase.hs


The diff was not included because it is too large.


View it on GitLab: https://gitlab.haskell.org/ghc/ghc/-/compare/ddbc64fe57fb86ce1ac3738d9b187c90262bd8bd...4477efff150d495cc7bd53d9805648f1829d94d7

-- 
View it on GitLab: https://gitlab.haskell.org/ghc/ghc/-/compare/ddbc64fe57fb86ce1ac3738d9b187c90262bd8bd...4477efff150d495cc7bd53d9805648f1829d94d7
You're receiving this email because of your account on gitlab.haskell.org.


-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mail.haskell.org/pipermail/ghc-commits/attachments/20230117/4bd0a822/attachment-0001.html>


More information about the ghc-commits mailing list