GitLab

Sven Tennie pushed to branch wip/romes/hadrian-cross-stage2-rebase_SVEN_FIXED at Glasgow Haskell Compiler / GHC

Commits:

e10dcd65

by Sven Tennie at 2025-10-12T10:24:56+00:00

T22859: Increase threadDelay for small machines

The previously used thread delay led to failures on my RISC-V test
setups.

d59ef6b6

by Hai / @BestYeen at 2025-10-14T21:51:14-04:00

Change Alex and Happy m4 scripts to display which version was found in the system, adapt small formatting details in Happy script to be more like the Alex script again.

c98abb6a

by Hai / @BestYeen at 2025-10-14T21:52:08-04:00

Update occurrences of return to pure and add a sample for redefining :m to mean :main

70ee825a

by Cheng Shao at 2025-10-14T21:52:50-04:00

testsuite: fix T3586 for non-SSE3 platforms

`T3586.hs` contains `-fvia-C -optc-msse3` which I think is a
best-effort basis to harvest the C compiler's auto vectorization
optimizations via the C backend back when the test was added. The
`-fvia-C` part is now a deprecated no-op because GHC can't fall back
to the C backend on a non-unregisterised build, and `-optc-msse3`
might actually cause the test to fail on non x86/x64 platforms, e.g.
recent builds of wasi-sdk would report `wasm32-wasi-clang: error:
unsupported option '-msse3' for target 'wasm32-unknown-wasi'`.

So this patch cleans up this historical cruft. `-fvia-C` is removed,
and `-optc-msse3` is only passed when cpuid contains `pni` (which
indicates support of SSE3).

4be32153

by Teo Camarasu at 2025-10-15T08:06:09-04:00

Add submodules for template-haskell-lift and template-haskell-quasiquoter

These two new boot libraries expose stable subsets of the
template-haskell interface.

This is an implemenation of the GHC proposal https://github.com/ghc-proposals/ghc-proposals/pull/696

Work towards #25262

0c00c9c3

by Ben Gamari at 2025-10-15T08:06:51-04:00

rts: Eliminate uses of implicit constant arrays

Folding of `const`-sized variable-length arrays to a constant-length
array is a gnu extension which clang complains about.

Closes #26502.

bf902a1d

by Fendor at 2025-10-15T16:00:59-04:00

Refactor distinct constructor tables map construction

Adds `GHC.Types.Unique.FM.alterUFM_L`, `GHC.Types.Unique.DFM.alterUDFM_L`
`GHC.Data.Word64Map.alterLookup` to support fusion of distinct
constructor data insertion and lookup during the construction of the `DataCon`
map in `GHC.Stg.Debug.numberDataCon`.

Co-authored-by: Fendor <fendor@posteo.de>
Co-authored-by: Finley McIlwaine <finleymcilwaine@gmail.com>

b3585ba1

by Fendor at 2025-10-15T16:00:59-04:00

Allow per constructor refinement of distinct-constructor-tables

Introduce `-fno-distinct-constructor-tables`. A distinct constructor table
configuration is built from the combination of flags given, in order. For
example, to only generate distinct constructor tables for a few specific
constructors and no others, just pass
`-fdistinct-constructor-tables-only=C1,...,CN`.
This flag can be supplied multiple times to extend the set of
constructors to generate a distinct info table for.

You can disable generation of distinct constructor tables for all
configurations by passing `-fno-distinct-constructor-tables`.

The various configurations of these flags is included in the `DynFlags`
fingerprints, which should result in the expected recompilation logic.

Adds a test that checks for distinct tables for various given or omitted
constructors.

Updates CountDepsAst and CountDepsParser tests to account for new dependencies.

Fixes #23703

Co-authored-by: Fendor <fendor@posteo.de>
Co-authored-by: Finley McIlwaine <finleymcilwaine@gmail.com>

e17dc695

by fendor at 2025-10-15T16:01:41-04:00

Fix typos in haddock documentation for stack annotation API

f85058d3

by Zubin Duggal at 2025-10-17T13:50:52+05:30

compiler: Attempt to systematize Unique tags by introducing an ADT for each different tag

Fixes #26264

Metric Decrease:
    T9233

c85c845d

by sheaf at 2025-10-17T22:35:32-04:00

Don't prematurely final-zonk PatSyn declarations

This commit makes GHC hold off on the final zonk for pattern synonym
declarations, in 'GHC.Tc.TyCl.PatSyn.tc_patsyn_finish'.
This accommodates the fact that pattern synonym declarations without a
type signature can contain unfilled metavariables, e.g. if the RHS of
the pattern synonym involves view-patterns whose type mentions promoted
(level 0) metavariables. Just like we do for ordinary function bindings,
we should allow these metavariables to be settled later, instead of
eagerly performing a final zonk-to-type.

Now, the final zonking-to-type for pattern synonyms is performed in
GHC.Tc.Module.zonkTcGblEnv.

Fixes #26465

ba3e5bdd

by Rodrigo Mesquita at 2025-10-18T16:57:18-04:00

Move code-gen aux symbols from ghc-internal to rts

These symbols were all previously defined in ghc-internal and made the
dependency structure awkward, where the rts may refer to some of these
symbols and had to work around that circular dependency the way
described in #26166.

Moreover, the code generator will produce code that uses these symbols!
Therefore, they should be available in the rts:

PRINCIPLE: If the code generator may produce code which uses this
symbol, then it should be defined in the rts rather than, say,
ghc-internal.

That said, the main motivation is towards fixing #26166.

Towards #26166. Pre-requisite of !14892

f31de2a9

by Ben Gamari at 2025-10-18T16:57:18-04:00

rts: Avoid static symbol references to ghc-internal

This resolves #26166, a bug due to new constraints placed by Apple's
linker on undefined references.

One source of such references in the RTS is the many symbols referenced
in ghc-internal. To mitigate #26166, we make these references dynamic,
as described in Note [RTS/ghc-internal interface].

Fixes #26166

Co-authored-by: Rodrigo Mesquita <rodrigo.m.mesquita@gmail.com>
Co-authored-by: Cheng Shao <terrorjack@type.dance>

43fdfddc

by Ben Gamari at 2025-10-18T16:57:18-04:00

compiler: Rename isMathFun -> isLibcFun

This set includes more than just math functions.

4ed5138f

by Ben Gamari at 2025-10-18T16:57:18-04:00

compiler: Add libc allocator functions to libc_funs

Prototypes for these are now visible from `Prim.h`, resulting in
multiple-declaration warnings in the unregisterised job.

9a0a076b

by Ben Gamari at 2025-10-18T16:57:18-04:00

rts: Minimize header dependencies of Prim.h

Otherwise we will end up with redundant and incompatible declarations
resulting in warnings during the unregisterised build.

26b8a414

by Diego Antonio Rosario Palomino at 2025-10-18T16:58:10-04:00

Cmm Parser: Fix incorrect example in comment

The Parser.y file contains a comment with an incorrect example of textual
Cmm (used in .cmm files). This commit updates the comment to ensure it
reflects valid textual Cmm syntax.

Fixes #26313

d4a9d6d6

by ARATA Mizuki at 2025-10-19T18:43:47+09:00

Handle implications between x86 feature flags

This includes:

* Multiple -msse* options can be specified
* -mavx implies -msse4.2
* -mavx2 implies -mavx
* -mfma implies -mavx
* -mavx512f implies -mavx2 and -mfma
* -mavx512{cd,er,pf} imply -mavx512f

Closes #24989

Co-authored-by: sheaf <sam.derbyshire@gmail.com>

c9b8465c

by Cheng Shao at 2025-10-20T10:16:00-04:00

wasm: workaround WebKit bug in dyld

This patch works around a WebKit bug and allows dyld to run on WebKit
based platforms as well. See added note for detailed explanation.

Co-authored-by: Codex <codex@openai.com>

91b6be10

by Julian Ospald at 2025-10-20T18:21:03-04:00

Improve error handling in 'getPackageArchives'

When the library dirs in the package conf files are not set up correctly,
the JS linker will happily ignore such packages and not link against them,
although they're part of the link plan.

Fixes #26383

6c5269da

by Sven Tennie at 2025-10-20T18:21:44-04:00

Align coding style

Improve readability by using the same style for all constructor calls in
this function.

3d305889

by Sven Tennie at 2025-10-20T18:21:44-04:00

Reduce complexity by removing joins with mempty

ldArgs, cArgs and cppArgs are all `mempty`. Thus concatenating them adds
nothing but some complexity while reading the code.

38d65187

by Matthew Pickering at 2025-10-21T13:12:20+01:00

Fix stack decoding when using profiled runtime

There are three fixes in this commit.

* We need to replicate the `InfoTable` and `InfoTableProf`
approach for the other stack constants (see the new Stack.ConstantsProf
file).

* Then we need to appropiately import the profiled or non-profiled
  versions.

* Finally, there was an incorrect addition in `stackFrameSize`. We need
  to cast after performing addition on words.

Fixes #26507

17231bfb
by fendor at 2025-10-21T13:12:20+01:00
```
Add regression test for #26507
```

4f5bf93b

by Simon Peyton Jones at 2025-10-25T04:05:34-04:00

Postscript to fix for #26255

This MR has comments only

6ef22fa0

by IC Rainbow at 2025-10-26T18:23:01-04:00

Add SIMD primops for bitwise logical operations

This adds 128-bit wide and/or/xor instructions for X86 NCG,
with both SSE and AVX encodings.

```
andFloatX4# :: FloatX4# -> FloatX4# -> FloatX4#     -- andps / vandps
andDoubleX2# :: DoubleX2# -> DoubleX2# -> DoubleX2# -- andpd / vandpd
andInt8X16# :: Int8X16# -> Int8X16# -> Int8X16#     -- pand / vpand
```

The new primops are available on ARM when using LLVM backend.

Tests added:
- simd015 (floats and doubles)
- simd016 (integers)
- simd017 (words)

Fixes #26417

fbdc623a

by sheaf at 2025-10-26T18:23:52-04:00

Add hints for unsolved HasField constraints

This commit adds hints and explanations for unsolved 'HasField'
constraints.

GHC will now provide additional explanations for an unsolved constraint
of the form 'HasField fld_name rec_ty fld_ty'; the details are laid out in
Note [Error messages for unsolved HasField constraints], but briefly:

  1. Provide similar name suggestions (e.g. mis-spelled field name)
     and import suggestions (record field not in scope).
     These result in actionable 'GhcHints', which is helpful to provide
     code actions in HLS.

  2. Explain why GHC did not solve the constraint, e.g.:

    - 'fld_name' is not a string literal (e.g. a type variable)
    - 'rec_ty' is a TyCon without any fields, e.g. 'Int' or 'Bool'.
    - 'fld_ty' contains existentials variables or foralls.
    - The record field is a pattern synonym field (GHC does not generate
      HasField instances for those).
    - 'HasField' is a custom 'TyCon', not actually the built-in
      'HasField' typeclass from 'GHC.Records'.

On the way, we slightly refactor the mechanisms for import suggestions
in GHC.Rename.Unbound. This is to account for the fact that, for
'HasField', we don't care whether the field is imported qualified or
unqualified. 'importSuggestions' was refactored, we now have
'sameQualImportSuggestions' and 'anyQualImportSuggestions'.

Fixes #18776 #22382 #26480

99d5707f

by sheaf at 2025-10-26T18:23:52-04:00

Rename PatSyn MatchContext to PatSynCtx to avoid punning

5dc2e9ea
by Julian Ospald at 2025-10-27T18:17:23-04:00
```
Skip uniques test if sources are not available
```

544b9ec9

by Vladislav Zavialov at 2025-10-27T18:18:06-04:00

Re-export GHC.Hs.Basic from GHC.Hs

Clean up some import sections in GHC by re-exporting GHC.Hs.Basic
from GHC.Hs.

643ce801

by Julian Ospald at 2025-10-28T18:18:55-04:00

rts: remove unneccesary cabal flags

We perform those checks via proper autoconf macros
instead that do the right thing and then add those
libs to the rts buildinfo.

d69ea8fe

by Vladislav Zavialov at 2025-10-28T18:19:37-04:00

Test case for #17705

Starting with GHC 9.12 (the first release to include 5745dbd3),
all examples in this ticket are handled as expected.

4038a28b
by Andreas Klebinger at 2025-10-30T12:38:52-04:00
```
Add a perf test for #26425
```

f997618e

by Andreas Klebinger at 2025-10-30T12:38:52-04:00

OccAnal: Be stricter for better compiler perf.

In particular we are now stricter:

* When combining usageDetails.
* When computing binder info.

In combineUsageDetails when combining the underlying adds we compute a
new `LocalOcc` for each entry by combining the two existing ones.

Rather than wait for those entries to be forced down the road we now
force them immediately. Speeding up T26425 by about 10% with little
effect on the common case.

We also force binders we put into the Core AST everywhere now.
Failure to do so risks leaking the occ env used to set the binders
OccInfo.

For T26425 compiler residency went down by a factor of ~10x.

Compile time also improved by a factor of ~1.6.

-------------------------
Metric Decrease:
    T18698a
    T26425
    T9233
-------------------------

5618645b

by Vladislav Zavialov at 2025-10-30T12:39:33-04:00

Fix namespace specifiers in subordinate exports (#12488)

This patch fixes an oversight in the `lookupChildrenExport` function that
caused explicit namespace specifiers of subordinate export items to be
ignored:

  module M (T (type A)) where   -- should be rejected
  data T = A

Based on the `IEWrappedName` data type, there are 5 cases to consider:
1. Unadorned name:   P(X)
2. Named default:    P(default X)
3. Pattern synonym:  P(pattern X)
4. Type name:        P(type X)
5. Data name:        P(data X)

Case 1 is already handled correctly; cases 2 and 3 are parse errors; and
it is cases 4 and 5 that we are concerned with in this patch.

Following the precedent established in `LookupExactName`, we introduce
a boolean flag in `LookupChildren` to control whether to look up in all
namespaces or in a specific one. If an export item is accompanied by an
explicit namespace specifier `type` or `data`, we restrict the lookup in
`lookupGRE` to a specific namespace.

The newly introduced diagnostic `TcRnExportedSubordinateNotFound`
provides error messages and suggestions more tailored to this context
than the previously used `reportUnboundName`.

f341ae29

by Matthew Pickering at 2025-10-31T15:31:53+01:00

Add missing req_interp modifier to T18441fail3 and T18441fail19

These tests require the interpreter but they were failing in a different
way with the javascript backend because the interpreter was disabled and
stderr is ignored by the test.

7581e62d
by Matthew Pickering at 2025-10-31T15:31:53+01:00
```
Use explicit syntax rather than pure
```

c532fb35

by Matthew Pickering at 2025-10-31T15:31:53+01:00

packaging: correctly propagate build/host/target to bindist configure script

At the moment the host and target which we will produce a compiler for
is fixed at the initial configure time. Therefore we need to persist
the choice made at this time into the installation bindist as well so we
look for the right tools, with the right prefixes at install time.

In the future, we want to provide a bit more control about what kind of
bindist we produce so the logic about what the host/target will have to
be written by hadrian rather than persisted by the configure script. In
particular with cross compilers we want to either build a normal stage 2
cross bindist or a stage 3 bindist, which creates a bindist which has a
native compiler for the target platform.

Fixes #21970

e5c4aa1b

by Matthew Pickering at 2025-10-31T15:31:53+01:00

hadrian: Fill in more of the default.host toolchain file

When you are building a cross compiler this file will be used to build
stage1 and it's libraries, so we need enough information here to work
accurately. There is still more work to be done (see for example, word
size is still fixed).

3d723daa

by Matthew Pickering at 2025-10-31T15:31:53+01:00

hadrian: Disable docs when cross compiling

Before there were a variety of ad-hoc places where doc building was
disabled when cross compiling.

* Some CI jobs sets --docs=none in gen_ci.hs
* Some CI jobs set --docs=none in .gitlab/ci.sh
* There was some logic in hadrian to not need the ["docs"] target when
  making a bindist.

Now the situation is simple:

* If you are cross compiling then defaultDocsTargets is empty by
  default.

In theory, there is no reason why we can't build documentation for cross
compiler bindists, but this is left to future work to generalise the
documentation building rules to allow this (#24289)

46386af6

by Matthew Pickering at 2025-10-31T16:25:24+01:00

hadrian: Build stage 2 cross compilers

* Most of hadrian is abstracted over the stage in order to remove the
  assumption that the target of all stages is the same platform. This
  allows the RTS to be built for two different targets for example.
* Abstracts the bindist creation logic to allow building either normal
  or cross bindists. Normal bindists use stage 1 libraries and a stage 2
  compiler. Cross bindists use stage 2 libararies and a stage 2
  compiler.
* hadrian: Make binary-dist-dir the default build target. This allows us
  to have the logic in one place about which libraries/stages to build
  with cross compilers. Fixes #24192

New hadrian target:

* `binary-dist-dir-cross`: Build a cross compiler bindist (compiler =
  stage 1, libraries = stage 2)

-------------------------
Metric Decrease:
    T10421a
    T10858
    T11195
    T11276
    T11374
    T11822
    T15630
    T17096
    T18478
    T20261
Metric Increase:
    parsing001
-------------------------

b0cd30b2

by Matthew Pickering at 2025-10-31T16:25:24+01:00

ci: Test cross bindists

We remove the special logic for testing in-tree cross
compilers and instead test cross compiler bindists, like we do for all
other platforms.

440eba45

by Matthew Pickering at 2025-10-31T16:25:24+01:00

ci: Javascript don't set CROSS_EMULATOR

There is no CROSS_EMULATOR needed to run javascript binaries, so we
don't set the CROSS_EMULATOR to some dummy value.

e16f1bb1

by Matthew Pickering at 2025-10-31T16:25:24+01:00

ci: Introduce CROSS_STAGE variable

In preparation for building and testing stage3 bindists we introduce the
CROSS_STAGE variable which is used by a CI job to determine what kind of
bindist the CI job should produce.

At the moment we are only using CROSS_STAGE=2 but in the future we will
have some jobs which set CROSS_STAGE=3 to produce native bindists for a
target, but produced by a cross compiler, which can be tested on by
another CI job on the native platform.

CROSS_STAGE=2: Build a normal cross compiler bindist
CROSS_STAGE=3: Build a stage 3 bindist, one which is a native compiler and library for the target

4d73c166

by Matthew Pickering at 2025-10-31T16:53:17+01:00

Split up system.config into host/target config files

There were a number of settings which were not applied per-stage, for
example if you specified `--ffi-include-dir` then that was applied to
both host and target. Now this will just be passed when building the
crosscompiler.

The solution for now is to separate these two files into host/target and
the host file contains very bare-bones . There isn't currently a way to
specify with configure anything in the host file, so if you are building
a cross-compiler and you need to do that, you have to modify the file
yourself.

f0797745
by Matthew Pickering at 2025-10-31T16:53:17+01:00
```
Fix location of emsdk-version
```
1bb16a1f
by Matthew Pickering at 2025-10-31T16:53:17+01:00
```
fix distrib/configure file
```
84037672
by Matthew Pickering at 2025-10-31T16:53:17+01:00
```
Fix hardcoded stage1
```
def77f8c
by Matthew Pickering at 2025-10-31T16:53:17+01:00
```
Don't recache
```

bcfd462d

by Matthew Pickering at 2025-10-31T16:53:17+01:00

hadrian: Make text_simdutf flavour transformer configurable per-stage

Before it was globally enabled, which was probably not what you want as
you don't need text-simd for your boot compiler nor your boot compiler
if you're building a cross-compiler.

This brings it into line with the other modifiers.. such as ghcProfiled
etc

Fixes #25302

fe720ff6

by Matthew Pickering at 2025-10-31T16:53:17+01:00

hadrian: Refactor system-cxx-std-lib rules0

I noticed a few things wrong with the hadrian rules for `system-cxx-std-lib` rules.

* For `text` there is an ad-hoc check to depend on `system-cxx-std-lib` outside of `configurePackage`.
* The `system-cxx-std-lib` dependency is not read from cabal files.
* Recache is not called on the packge database after the `.conf` file is generated, a more natural place for this rule is `registerRules`.

Treating this uniformly like other packages is complicated by it not having any source code or a cabal file. However we can do a bit better by reporting the dependency firstly in `PackageData` and then needing the `.conf` file in the same place as every other package in `configurePackage`.

Fixes #25303

d9b3aa9e
by Matthew Pickering at 2025-10-31T16:53:17+01:00
```
fixes for simdutf8
```
cbd932df
by Matthew Pickering at 2025-10-31T16:53:17+01:00
```
use building for target in llvm flavour transformer
```

b133917c

by Matthew Pickering at 2025-10-31T16:53:17+01:00

bindist: Pass path to package database we want to recache

This fixes recaching on cross compilers

25069175

by Matthew Pickering at 2025-10-31T16:53:17+01:00

testsuite: T9930fail now passes on javascript

I didn't investigate why, but the comment says it should be fixed by
building a stage2 cross compiler (and it is).

96c15474

by Matthew Pickering at 2025-10-31T16:53:17+01:00

hadrian: Fix predicate for building shared libraries in defaultLibraries

Obviously we should only attempt to build shared libraries if the target
supports building shared libraries.

16c94edd
by Matthew Pickering at 2025-10-31T17:17:16+01:00
```
Hard-code ways in settings
```

0ac506c8