[Git][ghc/ghc][wip/ghc-9.6-backports] 41 commits: Propagate long-distance info in generated code
Zubin (@wz1000)
gitlab at gitlab.haskell.org
Wed Sep 13 12:07:43 UTC 2023
Zubin pushed to branch wip/ghc-9.6-backports at Glasgow Haskell Compiler / GHC
Commits:
fedc7d73 by sheaf at 2023-09-13T17:18:14+05:30
Propagate long-distance info in generated code
When desugaring generated pattern matches, we skip pattern match checks.
However, this ended up also discarding long-distance information, which
might be needed for user-written sub-expressions.
Example:
```haskell
okay (GADT di) cd =
let sr_field :: ()
sr_field = case getFooBar di of { Foo -> () }
in case cd of { SomeRec _ -> SomeRec sr_field }
```
With sr_field a generated FunBind, we still want to propagate the outer
long-distance information from the GADT pattern match into the checks
for the user-written RHS of sr_field.
Fixes #23445
(cherry picked from commit fbc8e04e5d8fb05ff60568042802ab2fb34e1a70)
- - - - -
897b5689 by Richard Eisenberg at 2023-09-13T17:18:14+05:30
Don't suppress *all* Wanteds
Code in GHC.Tc.Errors.reportWanteds suppresses a Wanted if its
rewriters have unfilled coercion holes; see
Note [Wanteds rewrite Wanteds] in GHC.Tc.Types.Constraint.
But if we thereby suppress *all* errors that's really confusing,
and as #22707 shows, GHC goes on without even realising that the
program is broken. Disaster.
This MR arranges to un-suppress them all if they all get suppressed.
Close #22707
(cherry picked from commit 1ed573a53ee454db240b9fb1a17e28c97b6eb53a)
- - - - -
bd38bb14 by Luite Stegeman at 2023-09-13T17:18:14+05:30
Delete created temporary subdirectories at end of session.
This patch adds temporary subdirectories to the list of
paths do clean up at the end of the GHC session. This
fixes warnings about non-empty temporary directories.
Fixes #22952
(cherry picked from commit f97c7f6d96c58579d630bc883929afc3d45d5c2b)
- - - - -
0f4dfc0a by Matthew Pickering at 2023-09-13T17:18:14+05:30
Fix behaviour of -keep-tmp-files when used in OPTIONS_GHC pragma
This fixes the behaviour of -keep-tmp-files when used in an OPTIONS_GHC
pragma for files with module level scope.
Instead of simple not deleting the files, we also need to remove them
from the TmpFs so they are not deleted later on when all the other files
are deleted.
There are additional complications because you also need to remove the
directory where these files live from the TmpFs so we don't try to
delete those later either.
I added two tests.
1. Tests simply that -keep-tmp-files works at all with a single module
and --make mode.
2. The other tests that temporary files are deleted for other modules
which don't enable -keep-tmp-files.
Fixes #23339
(cherry picked from commit a24b83ddabac6b7eeb63db13884e4403f71375dd)
- - - - -
2423c854 by Matthew Pickering at 2023-09-13T17:18:14+05:30
withDeferredDiagnostics: When debugIsOn, write landmine into IORef to catch use-after-free.
Ticket #23305 reports an error where we were attempting to use the
logger which was created by withDeferredDiagnostics after its scope had
ended.
This problem would have been caught by this patch and a validate build:
```
+*** Exception: Use after free
+CallStack (from HasCallStack):
+ error, called at compiler/GHC/Driver/Make.hs:<line>:<column> in <package-id>:GHC.Driver.Make
```
This general issue is tracked by #20981
(cherry picked from commit dcf3288273d2418800e2dee97c937673a1d38a8f)
- - - - -
35e4c00c by Matthew Pickering at 2023-09-13T17:18:14+05:30
Don't return complete HscEnv from upsweep
By returning a complete HscEnv from upsweep the logger (as introduced by
withDeferredDiagnostics) was escaping the scope of
withDeferredDiagnostics and hence we were losing error messages.
This is reminiscent of #20981, which also talks about writing errors
into messages after their scope has ended.
See #23305 for details.
(cherry picked from commit 432c736c19446a011fca1f9485c67761c991bd42)
- - - - -
8ee3adf4 by Ryan Scott at 2023-09-13T17:18:14+05:30
Add regression test for #23143
!10541, the fix for #23323, also fixes #23143. Let's add a regression test to
ensure that it stays fixed.
Fixes #23143.
(cherry picked from commit 95b69cfb3d601eb3e6c5b1727c4cfef25ab87d68)
- - - - -
01cb005a by Simon Peyton Jones at 2023-09-13T17:18:14+05:30
Don't report redundant Givens from quantified constraints
This fixes #23323
See (RC4) in Note [Tracking redundant constraints]
(cherry picked from commit 2b0c9f5ef026df6dd2637aacce05a11d74146296)
- - - - -
bc04ca51 by Ben Gamari at 2023-09-13T17:18:14+05:30
nativeGen: Explicitly set flags of text sections on Windows
The binutils documentation (for COFF) claims,
> If no flags are specified, the default flags depend upon the section
> name. If the section name is not recognized, the default will be for the
> section to be loaded and writable.
We previously assumed that this would do the right thing for split
sections (e.g. a section named `.text$foo` would be correctly inferred
to be a text section). However, we have observed that this is not the
case (at least under the clang toolchain used on Windows): when
split-sections is enabled, text sections are treated by the assembler as
data (matching the "default" behavior specified by the documentation).
Avoid this by setting section flags explicitly. This should fix split
sections on Windows.
Fixes #22834.
(cherry picked from commit 3ece9856d157c85511d59f9f862ab351bbd9b38b)
- - - - -
3bc903b2 by Ben Gamari at 2023-09-13T17:18:14+05:30
nativeGen: Set explicit section types on all platforms
(cherry picked from commit db7f7240b53c01447e44d2790ee37eacaabfbcf3)
- - - - -
be64c6e6 by Ben Gamari at 2023-09-13T17:18:14+05:30
testsuite: Add tests for #23146
Both lifted and unlifted variants.
(cherry picked from commit 33cf4659f209ef8e97be188279216a2f4fe0cf51)
- - - - -
7f2f7ac1 by Ben Gamari at 2023-09-13T17:18:15+05:30
codeGen: Fix some Haddocks
(cherry picked from commit 76727617bccc88d1466ad6dc1442ab8ebb34f79a)
- - - - -
6eb8e32a by Ben Gamari at 2023-09-13T17:18:15+05:30
codeGen: Give proper LFInfo to datacon wrappers
As noted in `Note [Conveying CAF-info and LFInfo between modules]`,
when importing a binding from another module we must ensure that it gets
the appropriate `LambdaFormInfo` if it is in WHNF to ensure that
references to it are tagged correctly.
However, the implementation responsible for doing this,
`GHC.StgToCmm.Closure.mkLFImported`, only dealt with datacon workers and
not wrappers. This lead to the crash of this program in #23146:
module B where
type NP :: [UnliftedType] -> UnliftedType
data NP xs where
UNil :: NP '[]
module A where
import B
fieldsSam :: NP xs -> NP xs -> Bool
fieldsSam UNil UNil = True
x = fieldsSam UNil UNil
Due to its GADT nature, `UNil` produces a trivial wrapper
$WUNil :: NP '[]
$WUNil = UNil @'[] @~(<co:1>)
which is referenced in the RHS of `A.x`. Due to the above-mentioned bug
in `mkLFImported`, the references to `$WUNil` passed to `fieldsSam` were
not tagged. This is problematic as `fieldsSam` expected its arguments to
be tagged as they are unlifted.
The fix is straightforward: extend the logic in `mkLFImported` to cover
(nullary) datacon wrappers as well as workers. This is safe because we
know that the wrapper of a nullary datacon will be in WHNF, even if it
includes equalities evidence (since such equalities are not runtime
relevant).
Thanks to @MangoIV for the great ticket and @alt-romes for his
minimization and help debugging.
Fixes #23146.
(cherry picked from commit 33a8c348cae5fd800c015fd8c2230b8066c7c0a4)
- - - - -
81f2cceb by Rodrigo Mesquita at 2023-09-13T17:18:15+05:30
codeGen: Fix LFInfo of imported datacon wrappers
As noted in #23231 and in the previous commit, we were failing to give a
an LFInfo of LFCon to a nullary datacon wrapper from another module,
failing to properly tag pointers which ultimately led to the
segmentation fault in #23146.
On top of the previous commit which now considers wrappers where we
previously only considered workers, we change the order of the guards so
that we check for the arity of the binding before we check whether it is
a constructor. This allows us to
(1) Correctly assign `LFReEntrant` to imported wrappers whose worker was
nullary, which we previously would fail to do
(2) Remove the `isNullaryRepDataCon` predicate:
(a) which was previously wrong, since it considered wrappers whose
workers had zero-width arguments to be non-nullary and would fail to
give `LFCon` to them
(b) is now unnecessary, since arity == 0 guarantees
- that the worker takes no arguments at all
- and the wrapper takes no arguments and its RHS must be an
application of the worker to zero-width-args only.
- we lint these two items with an assertion that the datacon
`hasNoNonZeroWidthArgs`
We also update `isTagged` to use the new logic in determining the
LFInfos of imported Ids.
The creation of LFInfos for imported Ids and this detail are explained
in Note [The LFInfo of Imported Ids].
Note that before the patch to those issues we would already consider these
nullary wrappers to have `LFCon` lambda form info; but failed to re-construct
that information in `mkLFImported`
Closes #23231, #23146
(I've additionally batched some fixes to documentation I found while
investigating this issue)
(cherry picked from commit 2fc18e9e784ccc775db8b06a5d10986588cce74a)
- - - - -
9c99cd76 by Sebastian Graf at 2023-09-13T17:18:15+05:30
DmdAnal: Unleash demand signatures of free RULE and unfolding binders (#23208)
In #23208 we observed that the demand signature of a binder occuring in a RULE
wasn't unleashed, leading to a transitively used binder being discarded as
absent. The solution was to use the same code path that we already use for
handling exported bindings.
See the changes to `Note [Absence analysis for stable unfoldings and RULES]`
for more details.
I took the chance to factor out the old notion of a `PlusDmdArg` (a pair of a
`VarEnv Demand` and a `Divergence`) into `DmdEnv`, which fits nicely into our
existing framework. As a result, I had to touch quite a few places in the code.
This refactoring exposed a few small bugs around correct handling of bottoming
demand environments. As a result, some strictness signatures now mention uniques
that weren't there before which caused test output changes to T13143, T19969 and
T22112. But these tests compared whole -ddump-simpl listings which is a very
fragile thing to begin with. I changed what exactly they test for based on the
symptoms in the corresponding issues.
There is a single regression in T18894 because we are more conservative around
stable unfoldings now. Unfortunately it is not easily fixed; let's wait until
there is a concrete motivation before invest more time.
Fixes #23208.
(cherry picked from commit c30ac25f7dfaded58bb2ff85d4bffe662e4af8b1)
- - - - -
0d642d43 by Matthew Craven at 2023-09-13T17:18:15+05:30
StgToCmm: Upgrade -fcheck-prim-bounds behavior
Fixes #21054. Additionally, we can now check for range overlap
when generating Cmm for primops that use memcpy internally.
(cherry picked from commit 65a442fccd081d9370ae4ee4e74f116139b5c2c8)
- - - - -
fbeb839d by Ben Gamari at 2023-09-13T17:18:15+05:30
hadrian: Always canonicalize topDirectory
Hadrian's `topDirectory` is intended to provide an absolute path to the
root of the GHC tree. However, if the tree is reached via a symlink this
One question here is whether the `canonicalizePath` call is expensive
enough to warrant caching. In a quick microbenchmark I observed that
`canonicalizePath "."` takes around 10us per call; this seems
sufficiently low not to worry.
Alternatively, another approach here would have been to rather move the
canonicalization into `m4/fp_find_root.m4`. This would have avoided
repeated canonicalization but sadly path canonicalization is a hard
problem in POSIX shell.
Addresses #22451.
(cherry picked from commit 5efa9ca545d8d33b9be4fc0ba91af1db38f19276)
- - - - -
7ed005ca by aadaa_fgtaa at 2023-09-13T17:18:15+05:30
Optimise ELF linker (#23464)
- cache last elements of `relTable`, `relaTable` and `symbolTables` in `ocInit_ELF`
- cache shndx table in ObjectCode
- run `checkProddableBlock` only with debug rts
(cherry picked from commit b3e1436f968c0c36a27ea0339ee2554970b329fe)
- - - - -
7f9a10c7 by Ben Gamari at 2023-09-13T17:18:15+05:30
rts: Ensure that pinned allocations respect block size
Previously, it was possible for pinned, aligned allocation requests to
allocate beyond the end of the pinned accumulator block. Specifically,
we failed to account for the padding needed to achieve the requested
alignment in the "large object" check. With large alignment requests,
this can result in the allocator using the capability's pinned object
accumulator block to service a request which is larger than
`PINNED_EMPTY_SIZE`.
To fix this we reorganize `allocatePinned` to consistently account for
the alignment padding in all large object checks. This is a bit subtle
as we must handle the case of a small allocation request filling the
accumulator block, as well as large requests.
Fixes #23400.
(cherry picked from commit fd8c57694a00f6359bd66365f1284388c869ac60)
- - - - -
1f788005 by Ben Gamari at 2023-09-13T17:18:15+05:30
testsuite: Add test for #23400
(cherry picked from commit 98185d5212fb0464dcbcca0ca2c33326a7a002e8)
- - - - -
1ad2e1cd by Ben Gamari at 2023-09-13T17:18:15+05:30
base: Fix incorrect CPP guard
This was guarded on `darwin_HOST_OS` instead of `defined(darwin_HOST_OS)`.
(cherry picked from commit d7ef1704aeba451bd3e0efbdaaab2638ee1f0bc8)
- - - - -
8dae53e2 by Ben Gamari at 2023-09-13T17:18:15+05:30
rts/Trace: Ensure that debugTrace arguments are used
As debugTrace is a macro we must take care to ensure that
the fact is clear to the compiler lest we see warnings.
(cherry picked from commit 7c7d1f66d35f73a2faa898a33aa80cd276159dc2)
- - - - -
622b09a8 by Ben Gamari at 2023-09-13T17:18:15+05:30
rts: Various warnings fixes
(cherry picked from commit cb92051e3d85575ff6abd753c9b135930cc50cf8)
- - - - -
9cdd8f41 by Ben Gamari at 2023-09-13T17:18:15+05:30
hadrian: Ignore warnings in unix and semaphore-compat
(cherry picked from commit dec81dd1fd0475dde4929baae625d155387300bb)
- - - - -
686a86b0 by Moisés Ackerman at 2023-09-13T17:35:18+05:30
Add failing test case for #23492
(cherry picked from commit 6074cc3cda9b9836c784942a1aa7f766fb142787)
- - - - -
469da90f by Moisés Ackerman at 2023-09-13T17:35:18+05:30
Use generated src span for catch-all case of record selector functions
This fixes #23492. The problem was that we used the real source span
of the field declaration for the generated catch-all case in the
selector function, in particular in the generated call to
`recSelError`, which meant it was included in the HIE output. Using
`generatedSrcSpan` instead means that it is not included.
(cherry picked from commit 356a269258a50bf67811fe0edb193fc9f82dfad1)
- - - - -
ae8571ff by Matthew Pickering at 2023-09-13T17:35:18+05:30
Add -fpolymorphic-specialisation flag (off by default at all optimisation levels)
Polymorphic specialisation has led to a number of hard to diagnose
incorrect runtime result bugs (see #23469, #23109, #21229, #23445) so
this commit introduces a flag `-fpolymorhphic-specialisation` which
allows users to turn on this experimental optimisation if they are
willing to buy into things going very wrong.
Ticket #23469
(cherry picked from commit 9f01d14b5bc1c73828b2b061206c45b84353620e)
- - - - -
e90957af by Bryan Richter at 2023-09-13T17:35:18+05:30
Add missing void prototypes to rts functions
See #23561.
(cherry picked from commit 82ac6bf113526f61913943b911089534705984fb)
- - - - -
c1f910d0 by Ben Gamari at 2023-09-13T17:35:18+05:30
Define FFI_GO_CLOSURES
The libffi shipped with Apple's XCode toolchain does not contain a
definition of the FFI_GO_CLOSURES macro, despite containing references
to said macro. Work around this by defining the macro, following the
model of a similar workaround in OpenJDK [1].
[1] https://github.com/openjdk/jdk17u-dev/pull/741/files
(cherry picked from commit 8b35e8caafeeccbf06b7faa70e807028a3f0ff43)
- - - - -
36dc5121 by Ben Gamari at 2023-09-13T17:35:18+05:30
hadrian: Ensure that way-flags are passed to CC
Previously the way-specific compilation flags (e.g. `-DDEBUG`,
`-DTHREADED_RTS`) would not be passed to the CC invocations. This meant
that C dependency files would not correctly reflect
dependencies predicated on the way, resulting in the rather
painful #23554.
Closes #23554.
(cherry picked from commit cca74dab6809f8cf7ffc2ec9df689e06aa425110)
- - - - -
b6bf7b43 by Krzysztof Gogolewski at 2023-09-13T17:35:18+05:30
Fix #23567, a specializer bug
Found by Simon in https://gitlab.haskell.org/ghc/ghc/-/issues/23567#note_507834
The testcase isn't ideal because it doesn't detect the bug in master,
unless doNotUnbox is removed as in
https://gitlab.haskell.org/ghc/ghc/-/issues/23567#note_507692.
But I have confirmed that with that modification, it fails before
and passes afterwards.
(cherry picked from commit bf9b9de0685e23c191722dfdb78d28b44f1cba05)
- - - - -
2086ffb5 by Dave Barton at 2023-09-13T17:35:18+05:30
Fix some broken links and typos
(cherry picked from commit 4457da2a7dba97ab2cd2f64bb338c904bb614244)
- - - - -
62d117c3 by Bodigrim at 2023-09-13T17:35:18+05:30
Add since annotations for Data.Foldable1
(cherry picked from commit 054261dd319b505392458da7745e768847015887)
- - - - -
1aef9974 by Ben Gamari at 2023-09-13T17:35:18+05:30
rts/RtsSymbols: Add AArch64 outline atomic operations
Fixes #22012 by adding the symbols described in
https://github.com/llvm/llvm-project/blob/main/llvm/docs/Atomics.rst#libcalls-atomic.
Ultimately this would be better addressed by #22011, but this is a first
step in the right direction and fixes the immediate symptom.
Note that we dropped the `__arch64_cas16` operations as these provided
by all platforms's compilers. Also, we don't link directly against the
libgcc/compiler-rt definitions but rather provide our own wrappers to
work around broken toolchains (e.g. https://bugs.gentoo.org/868018).
Generated via https://gitlab.haskell.org/ghc/ghc/-/snippets/5733.
(cherry picked from commit 1aa5733a4480420fdc146322d86dd143321a3da6)
- - - - -
d09e1901 by Matthew Pickering at 2023-09-13T17:35:18+05:30
driver: Fix -S with .cmm files
There was an oversight in the driver which assumed that you would always
produce a `.o` file when compiling a .cmm file.
Fixes #23610
(cherry picked from commit 76983a0dca64dfb7e94aea0c4f494921f8513b41)
- - - - -
380c8328 by sheaf at 2023-09-13T17:35:19+05:30
Valid hole fits: don't panic on a Given
The function GHC.Tc.Errors.validHoleFits would end up panicking when
encountering a Given constraint. To fix this, it suffices to filter out
the Givens before continuing.
Fixes #22684
(cherry picked from commit 630e302617a4a3e00d86d0650cb86fa9e6913e44)
- - - - -
e7406e9e by Matthew Pickering at 2023-09-13T17:35:19+05:30
simplifier: Correct InScopeSet in rule matching
The in-scope set passedto the `exprIsLambda_maybe` call lacked all the
in-scope binders. @simonpj suggests this fix where we augment the
in-scope set with the free variables of expression which fixes this
failure mode in quite a direct way.
Fixes #23630
(cherry picked from commit 4f5538a8e2a8b9bc490bcd098fa38f6f7e9f4d73)
- - - - -
db6198a0 by Ben Gamari at 2023-09-13T17:35:19+05:30
rts/win32: Ensure reliability of IO manager shutdown
When the Win32 threaded IO manager shuts down, `ioManagerDie` sends an
`IO_MANAGER_DIE` event to the IO manager thread using the
`io_manager_event` event object. Finally, it will closes the event object,
and invalidate `io_manager_event`.
Previously, `readIOManagerEvent` would see that `io_manager_event` is
invalid and return `0`, suggesting that everything is right with the
world. This meant that if `ioManagerDie` invalidated the handle before
the event manager was blocked on the event we would end up in a
situation where the event manager would never realize it was asked to
shut down.
Fix this by ensuring that `readIOManagerEvent` instead returns
`IO_MANAGER_DIE` when we detect that the event object has been
invalidated by `ioManagerDie`.
Fixes #23691.
(cherry picked from commit 01db1117e18f140987f608a78f3e929242d6f00c)
- - - - -
48917633 by Ben Gamari at 2023-09-13T17:35:19+05:30
codeGen: Ensure that TSAN is aware of writeArray# write barriers
By using a proper release store instead of a fence.
(cherry picked from commit aca20a5d4fde1c6429c887624bb95c9b54b7af73)
- - - - -
aa375afc by Ben Gamari at 2023-09-13T17:35:19+05:30
codeGen: Ensure that array reads have necessary barriers
This was the cause of #23541.
(cherry picked from commit 453c0531f2edf49b75c73bc45944600d8d7bf767)
- - - - -
c728db01 by Ben Gamari at 2023-09-13T17:35:19+05:30
linker/PEi386: Don't sign-extend symbol section number
Previously we incorrectly interpreted PE section numbers as signed
values. However, this isn't the case; rather, it's an unsigned 16-bit number
with a few special bit-patterns (0xffff and 0xfffe). This resulted in #22941
as the linker would conclude that the sections were invalid.
Fixing this required quite a bit of refactoring.
Closes #22941.
(cherry picked from commit 0eb54c050e46f447224167166dd6d2805ca8cdf5)
- - - - -
30 changed files:
- .gitlab-ci.yml
- .gitlab/gen_ci.hs
- .gitlab/jobs.yaml
- compiler/GHC/Cmm/CLabel.hs
- compiler/GHC/CmmToAsm/Ppr.hs
- compiler/GHC/Core/DataCon.hs
- compiler/GHC/Core/Opt/DmdAnal.hs
- compiler/GHC/Core/Opt/SpecConstr.hs
- compiler/GHC/Core/Opt/Specialise.hs
- compiler/GHC/Core/Rules.hs
- compiler/GHC/Core/Tidy.hs
- compiler/GHC/CoreToStg/Prep.hs
- compiler/GHC/Driver/CodeOutput.hs
- compiler/GHC/Driver/Flags.hs
- compiler/GHC/Driver/Make.hs
- compiler/GHC/Driver/Pipeline.hs
- compiler/GHC/Driver/Session.hs
- compiler/GHC/Hs/Utils.hs
- compiler/GHC/HsToCore/Match.hs
- compiler/GHC/HsToCore/Pmc.hs
- compiler/GHC/Linker/Static.hs
- compiler/GHC/Stg/InferTags/Rewrite.hs
- compiler/GHC/Stg/Syntax.hs
- compiler/GHC/Stg/Unarise.hs
- compiler/GHC/StgToCmm/Closure.hs
- compiler/GHC/StgToCmm/Env.hs
- compiler/GHC/StgToCmm/Foreign.hs
- compiler/GHC/StgToCmm/Monad.hs
- compiler/GHC/StgToCmm/Prim.hs
- compiler/GHC/StgToCmm/Types.hs
The diff was not included because it is too large.
View it on GitLab: https://gitlab.haskell.org/ghc/ghc/-/compare/dd3f569a35d31361707fdf75f383b7a53968e032...c728db017fa6bb1813deb3511e2d0ca97bbb3ea3
--
View it on GitLab: https://gitlab.haskell.org/ghc/ghc/-/compare/dd3f569a35d31361707fdf75f383b7a53968e032...c728db017fa6bb1813deb3511e2d0ca97bbb3ea3
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/20230913/f311759e/attachment-0001.html>
More information about the ghc-commits
mailing list