[Git][ghc/ghc][wip/ghc-9.6-backports] 39 commits: Show an error when we cannot default a concrete tyvar
Zubin (@wz1000)
gitlab at gitlab.haskell.org
Mon Sep 18 06:46:54 UTC 2023
Zubin pushed to branch wip/ghc-9.6-backports at Glasgow Haskell Compiler / GHC
Commits:
938a6614 by Krzysztof Gogolewski at 2023-09-18T10:07:51+05:30
Show an error when we cannot default a concrete tyvar
Fixes #23153
(cherry picked from commit 0da18eb79540181ae9835e73d52ba47ec79fff6b)
(cherry picked from commit 39574e3402ac33eb346e508da2667b9f337a590f)
- - - - -
f454428c by sheaf at 2023-09-18T10:07:51+05:30
Handle ConcreteTvs in inferResultToType
This patch fixes two issues.
1. inferResultToType was discarding the ir_frr information, which meant
some metavariables ended up being MetaTvs instead of ConcreteTvs.
This function now creates new ConcreteTvs as necessary, instead of
always creating MetaTvs.
2. startSolvingByUnification can make some type variables concrete.
However, it didn't return an updated type, so callers of this
function, if they don't zonk, might miss this and accidentally
perform a double update of a metavariable.
We now return the updated type from this function, which avoids
this issue.
Fixes #23154
(cherry picked from commit 9ab9b30ec1affe22b188f9a6637ac3bdea75bdba)
- - - - -
896fb39b by Krzysztof Gogolewski at 2023-09-18T10:07:51+05:30
Use tcInferFRR to prevent bad generalisation
Fixes #23176
(cherry picked from commit 4b89bb54a1d1d6a7b30a6bbfd21eed5d85506813)
- - - - -
5a2ac04e by sheaf at 2023-09-18T10:07:51+05:30
exactprint: silence incomplete record update warnings
(cherry picked from commit 860f6269bc016e11400b7e3176a5ea6dfe291a46)
- - - - -
45fb59a5 by sheaf at 2023-09-18T10:07:51+05:30
Reinstate untouchable variable error messages
This extra bit of information was accidentally being discarded after
a refactoring of the way we reported problems when unifying a type
variable with another type. This patch rectifies that.
(cherry picked from commit 2b55cb5f33666a71eaac7968c59e483860112e5c)
- - - - -
31d9a9fb by sheaf at 2023-09-18T10:07:51+05:30
Re-instate -Wincomplete-record-updates
Commit e74fc066 refactored the handling of record updates to use
the HsExpanded mechanism. This meant that the pattern matching inherent
to a record update was considered to be "generated code", and thus we
stopped emitting "incomplete record update" warnings entirely.
This commit changes the "data Origin = Source | Generated" datatype,
adding a field to the Generated constructor to indicate whether we
still want to perform pattern-match checking. We also have to do a bit
of plumbing with HsCase, to record that the HsCase arose from an
HsExpansion of a RecUpd, so that the error message continues to mention
record updates as opposed to a generic "incomplete pattern matches in case"
error.
Finally, this patch also changes the way we handle inaccessible code
warnings. Commit e74fc066 was also a regression in this regard, as we
were emitting "inaccessible code" warnings for case statements spuriously
generated when desugaring a record update (remember: the desugaring mechanism
happens before typechecking; it thus can't take into account e.g. GADT information
in order to decide which constructors to include in the RHS of the desugaring
of the record update).
We fix this by changing the mechanism through which we disable inaccessible
code warnings: we now check whether we are in generated code in
GHC.Tc.Utils.TcMType.newImplication in order to determine whether to
emit inaccessible code warnings.
Fixes #23520
Updates haddock submodule, to avoid incomplete record update warnings
(cherry picked from commit df706de378e3415a3972ddd14863f54fc7162dc7)
- - - - -
ee3921e9 by sheaf at 2023-09-18T10:07:51+05:30
Propagate long-distance information in do-notation
The preceding commit re-enabled pattern-match checking inside record
updates. This revealed that #21360 was in fact NOT fixed by e74fc066.
This commit makes sure we correctly propagate long-distance information
in do blocks, e.g. in
```haskell
data T = A { fld :: Int } | B
f :: T -> Maybe T
f r = do
a at A{} <- Just r
Just $ case a of { A _ -> A 9 }
```
we need to propagate the fact that "a" is headed by the constructor "A"
to see that the case expression "case a of { A _ -> A 9 }" cannot fail.
Fixes #21360
(cherry picked from commit 1d05971e24f6cb1120789d1e1ab4f086eebd504a)
- - - - -
4287c945 by sheaf at 2023-09-18T10:07:51+05:30
Skip PMC for boring patterns
Some patterns introduce no new information to the pattern-match
checker (such as plain variable or wildcard patterns). We can thus
skip doing any pattern-match checking on them when the sole purpose
for doing so was introducing new long-distance information.
See Note [Boring patterns] in GHC.Hs.Pat.
Doing this avoids regressing in performance now that we do additional
pattern-match checking inside do notation.
(cherry picked from commit bea0e323c09e9e4b841a37aacd6b67e87a85e7cb)
- - - - -
8182a259 by Sven Tennie at 2023-09-18T10:16:03+05:30
Add test for %mulmayoflo primop
The test expects a perfect implementation with no false positives.
(cherry picked from commit a36f9dc94823c75fb789710bc67b92e87a630440)
- - - - -
c81bb2c1 by Ben Gamari at 2023-09-18T10:16:08+05:30
testsuite: Mark MulMayOflo_full as broken rather than skipping
To ensure that we don't accidentally fix it.
See #23742.
(cherry picked from commit fd7ce39c70f8922e26b8be8a5fc4d6797987f66f)
- - - - -
271a9b34 by Ben Gamari at 2023-09-18T10:16:20+05:30
nativeGen/AArch64: Fix sign extension in MulMayOflo
Previously the 32-bit implementations of MulMayOflo would use the
a non-sensical sign-extension mode. Rewrite these to reflect what gcc 11
produces. Also similarly rework the 16- and 8-bit cases.
This now passes the MulMayOflo tests in ghc/test-primops> in all four
widths, including the precision tests.
Fixes #23721.
(cherry picked from commit 824092f28f52d32b6ea3cd26e1e576524ee24969)
- - - - -
7adb4da4 by Ben Gamari at 2023-09-18T10:33:47+05:30
compiler: Fingerprint more code generation flags
Previously our recompilation check was quite inconsistent in its
coverage of non-optimisation code generation flags. Specifically, we
failed to account for most flags that would affect the behavior of
generated code in ways that might affect the result of a program's
execution (e.g. `-feager-blackholing`, `-fstrict-dicts`)
Closes #23369.
(cherry picked from commit d1c92bf3b4b0b07a6a652f8fc31fd7b62465bf71)
- - - - -
e4e8c5f8 by Finley McIlwaine at 2023-09-18T10:33:47+05:30
Add -dipe-stats flag
This is useful for seeing which info tables have information.
(cherry picked from commit cc52c358316ac8210f80da80db6b0c620dd5bdc3)
- - - - -
1cd1987e by Finley McIlwaine at 2023-09-18T10:40:51+05:30
Add -finfo-table-map-with-fallback -finfo-table-map-with-stack
The -fno-info-table-map-with-stack flag omits STACK info tables from the info
table map, and the -fno-info-table-map-with-fallback flag omits info tables
with defaulted source locations from the map. In a test on the Agda codebase
the build results were about 7% smaller when both of those types of tables
were omitted.
Adds a test that verifies that passing each combination of these flags
results in the correct output for -dipe-stats, which is disabled for the js
backend since profiling is not implemented.
This commit also refactors a lot of the logic around extracting info tables
from the Cmm results and building the info table map.
This commit also fixes some issues in the users guide rst source to fix
warnings that were noticed while debugging the documentation for these flags.
Fixes #23702
(cherry picked from commit 261c4acbfdaf5babfc57ab0cef211edb66153fb1)
- - - - -
3f17c148 by Finley McIlwaine at 2023-09-18T10:48:17+05:30
Refactor estimation of stack info table provenance
This commit greatly refactors the way we compute estimated provenance for stack
info tables. Previously, this process was done using an entirely separate traversal
of the whole Cmm code stream to build the map from info tables to source locations.
The separate traversal is now fused with the Cmm code generation pipeline in
GHC.Driver.Main.
This results in very significant code generation speed ups when -finfo-table-map is
enabled. In testing, this patch reduces code generation times by almost 30% with
-finfo-table-map and -O0, and 60% with -finfo-table-map and -O1 or -O2 .
Fixes #23103
(cherry picked from commit d99c816f7b5727a3f344960e02a1932187ea093f)
- - - - -
98920bc8 by Finley McIlwaine at 2023-09-18T10:48:28+05:30
Add a test checking overhead of -finfo-table-map
We want to make sure we don't end up with poor codegen performance resulting from
-finfo-table-map again as in #23103. This test adds a performance test tracking
total allocations while compiling ExactPrint with -finfo-table-map.
(cherry picked from commit d3e0124c1157a4a423d86a1dc1d7e82c6d32ef06)
- - - - -
310c9d00 by Andreas Klebinger at 2023-09-18T10:54:06+05:30
Arm: Fix lack of zero-extension for 8/16 bit add/sub with immediate.
For 32/64bit we can avoid explicit extension/zeroing as the instructions
set the full width of the registers.
When doing 16/8bit computation we have to put a bit more work in so we
can't use the fast path.
Fixes #23749 for 9.4.
(cherry picked from commit 0bb44f695bd008f03644e3d306566c50c5bd528c)
- - - - -
4b14a2d1 by Ryan Scott at 2023-09-18T10:55:15+05:30
Restore mingwex dependency on Windows
This partially reverts some of the changes in !9475 to make `base` and
`ghc-prim` depend on the `mingwex` library on Windows. It also restores the
RTS's stubs for `mingwex`-specific symbols such as `_lock_file`.
This is done because the C runtime provides `libmingwex` nowadays, and
moreoever, not linking against `mingwex` requires downstream users to link
against it explicitly in difficult-to-predict circumstances. Better to always
link against `mingwex` and prevent users from having to do the guesswork
themselves.
See https://gitlab.haskell.org/ghc/ghc/-/merge_requests/10360#note_495873 for
the discussion that led to this.
(cherry picked from commit 2b1a4abe3f5935ca58c84c6073e6bdfa5160832f)
- - - - -
d97180eb by Ryan Scott at 2023-09-18T10:55:26+05:30
RtsSymbols.c: Remove mingwex symbol stubs
As of !9475, the RTS now links against `ucrt` instead of `msvcrt` on Windows,
which means that the RTS no longer needs to declare stubs for the `__mingw_*`
family of symbols. Let's remove these stubs to avoid confusion.
Fixes #23309.
(cherry picked from commit 289547580b6f2808ee123f106c3118b716486d5b)
- - - - -
ac3f91e7 by Jaro Reinders at 2023-09-18T11:02:55+05:30
Make STG rewriter produce updatable closures
(cherry picked from commit 3930d793901d72f42b1535c85b746f32d5f3b677)
- - - - -
87eaf730 by Ben Gamari at 2023-09-18T11:04:05+05:30
users-guide: Support both distutils and packaging
As noted in #23818, some old distributions (e.g. Debian 9) only include
`distutils` while newer distributions only include `packaging`.
Fixes #23818.
(cherry picked from commit d814bda97994df01139c2a9bcde915dc86ef2927)
- - - - -
21c30e96 by Ben Gamari at 2023-09-18T11:04:31+05:30
users-guide: Ensure extlinks is compatible with Sphinx <4
The semantics of the `extlinks` attribute annoyingly changed in Sphinx
4. Reflect this in our configuration. See #22690.
Fixes #23807.
(cherry picked from commit 1726db3f39f1c41b92b1bdf45e9dc054b401e782)
- - - - -
f636eb0c by Krzysztof Gogolewski at 2023-09-18T11:06:10+05:30
Fix MultiWayIf linearity checking (#23814)
Co-authored-by: Thomas BAGREL <thomas.bagrel at tweag.io>
(cherry picked from commit edd8bc43566b3f002758e5d08c399b6f4c3d7443)
- - - - -
f590f472 by Gergő Érdi at 2023-09-18T11:06:43+05:30
If we have multiple defaulting plugins, then we should zonk in between them
after any defaulting has taken place, to avoid a defaulting plugin seeing
a metavariable that has already been filled.
Fixes #23821.
(cherry picked from commit 1d92f2dff6d1a170a44488d73cef81292591d120)
- - - - -
e90574bf by Gergő Érdi at 2023-09-18T11:07:59+05:30
Improvements to the documentation of defaulting plugins
Based on @simonpj's draft and comments in !11117
(cherry picked from commit eaee4d296a0782c1acfde610ed3f0a7c7668c06c)
- - - - -
63f99374 by Matthew Pickering at 2023-09-18T11:08:49+05:30
driver: Check transitive closure of haskell package dependencies when deciding whether to relink
We were previously just checking whether direct package dependencies had
been modified. This caused issues when compiling without optimisations
as we wouldn't relink the direct dependency if one of its dependenices
changed.
Fixes #23724
(cherry picked from commit 291d81aef8083290da0d2ce430fbc5e5a33bdb6e)
- - - - -
5633affc by Ben Gamari at 2023-09-18T11:09:26+05:30
rts: Fix invalid symbol type
I suspect this code is dead since we haven't observed this failing
despite the obviously incorrect macro name.
(cherry picked from commit 9861f787a8323d03311e30851b10fdf100717afb)
- - - - -
df2d863d by Ben Gamari at 2023-09-18T11:10:15+05:30
testsuite: Add simple test exercising C11 atomics in GHCi
See #22012.
(cherry picked from commit 03ed6a9a634fd6c3ef35e9c5428b4a911e3f0add)
- - - - -
40cae442 by Ben Gamari at 2023-09-18T11:12:52+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)
- - - - -
1153aeac by Matthew Craven at 2023-09-18T11:17:07+05:30
Unarise: Split Rubbish literals in function args
Fixes #23914. Also adds a check to STG lint that
these args are properly unary or nullary after unarisation
(cherry picked from commit da30f0beb9e1820500382da02ffce96da959fa84)
- - - - -
b061cc4b by Simon Peyton Jones at 2023-09-18T11:21:32+05:30
Tiny refactor
canEtaReduceToArity was only called internally, and always with
two arguments equal to zero. This patch just specialises the
function, and renames it to cantEtaReduceFun.
No change in behaviour.
(cherry picked from commit 236a134eab4c0a3aae30752a3d580c083f4e6b57)
- - - - -
2d4fe6cb by Simon Peyton Jones at 2023-09-18T11:22:07+05:30
Fix eta reduction
Issue #23922 showed that GHC was bogusly eta-reducing a join point.
We should never eta-reduce (\x -> j x) to j, if j is a join point.
It is extremly difficult to trigger this bug. It took me 45 mins of
trying to make a small tests case, here immortalised as T23922a.
(cherry picked from commit 6840012e5bb8f5c13e4bf7a4e4cbba0b06420aaa)
- - - - -
c73198a1 by Andreas Klebinger at 2023-09-18T11:22:24+05:30
Profiling: Properly escape characters when using `-pj`.
There are some ways in which unusual characters like quotes or others
can make it into cost centre names. So properly escape these.
Fixes #23924
(cherry picked from commit e5c00092a13f1a8cf53df2469e027012743cf59a)
- - - - -
6a8b3fec by Krzysztof Gogolewski at 2023-09-18T11:25:59+05:30
Fix wrong role in mkSelCo_maybe
In the Lint failure in #23938, we start with a coercion Refl :: T a ~R T a,
and call mkSelCo (SelTyCon 1 nominal) Refl.
The function incorrectly returned Refl :: a ~R a. The returned role
should be nominal, according to the SelCo rule:
co : (T s1..sn) ~r0 (T t1..tn)
r = tyConRole tc r0 i
----------------------------------
SelCo (SelTyCon i r) : si ~r ti
In this test case, r is nominal while r0 is representational.
(cherry picked from commit e0aa8c6e3a8b6004eca9349e5b705b8a767050aa)
- - - - -
6a7e4af8 by Ben Gamari at 2023-09-18T11:27:16+05:30
base: Advertise linear time of readFloat
As noted in #23538, `readFloat` has runtime that scales nonlinearly in
the size of its input. Consequently, its use on untrusted input can
be exploited as a denial-of-service vector. Point this out and suggest
use of `read` instead.
See #23538.
(cherry picked from commit b33113c86ce5888ff5edfd6d3dd95772d3c8abce)
- - - - -
abe0c868 by Sylvain Henry at 2023-09-18T11:28:18+05:30
Add missing int64/word64-to-double/float rules (#23907)
CLC proposal: https://github.com/haskell/core-libraries-committee/issues/203
(cherry picked from commit 5126a2fef0385e206643b6af0543d10ff0c219d8)
- - - - -
d3be8e6e by Matthew Pickering at 2023-09-18T11:34:25+05:30
Build vanilla alpine bindists
We currently attempt to build and distribute fully static alpine
bindists (ones which could be used on any linux platform) but most
people who use the alpine bindists want to use alpine to build their own
static applications (for which a fully static bindist is not necessary).
We should build and distribute these bindists for these users whilst the
fully-static bindist is still unusable.
Fixes #23349
(cherry picked from commit 29be39ba3f187279b19cf451f2d8f58822edab4f)
- - - - -
bb16aae9 by Zubin Duggal at 2023-09-18T11:41:05+05:30
Bump filepath submodule to 1.4.100.4
Bump bytestring submodule to 0.11.5.2
- - - - -
6f840efd by Zubin Duggal at 2023-09-18T12:16:04+05:30
Prepare release 9.6.3
- - - - -
30 changed files:
- .gitlab/gen_ci.hs
- .gitlab/jobs.yaml
- .gitlab/rel_eng/fetch-gitlab-artifacts/fetch_gitlab.py
- compiler/GHC/Cmm.hs
- compiler/GHC/Cmm/MachOp.hs
- compiler/GHC/CmmToAsm/AArch64/CodeGen.hs
- compiler/GHC/Core/Coercion.hs
- compiler/GHC/Core/Opt/Arity.hs
- compiler/GHC/Driver/Config/StgToCmm.hs
- compiler/GHC/Driver/Flags.hs
- compiler/GHC/Driver/GenerateCgIPEStub.hs
- compiler/GHC/Driver/Main.hs
- compiler/GHC/Driver/Pipeline.hs
- compiler/GHC/Driver/Session.hs
- compiler/GHC/Hs/Expr.hs
- compiler/GHC/Hs/Pat.hs
- compiler/GHC/Hs/Utils.hs
- compiler/GHC/HsToCore/Arrows.hs
- compiler/GHC/HsToCore/Errors/Ppr.hs
- compiler/GHC/HsToCore/Errors/Types.hs
- compiler/GHC/HsToCore/Expr.hs
- compiler/GHC/HsToCore/GuardedRHSs.hs
- compiler/GHC/HsToCore/ListComp.hs
- compiler/GHC/HsToCore/Match.hs
- compiler/GHC/HsToCore/Match.hs-boot
- compiler/GHC/HsToCore/Match/Constructor.hs
- compiler/GHC/HsToCore/Monad.hs
- compiler/GHC/HsToCore/Pmc.hs
- compiler/GHC/HsToCore/Pmc/Utils.hs
- compiler/GHC/HsToCore/Utils.hs
The diff was not included because it is too large.
View it on GitLab: https://gitlab.haskell.org/ghc/ghc/-/compare/f939a7f7031695843aaff33039449b5b1509ffb5...6f840efd4ed336cf9850781d9a8a829fff392dc1
--
View it on GitLab: https://gitlab.haskell.org/ghc/ghc/-/compare/f939a7f7031695843aaff33039449b5b1509ffb5...6f840efd4ed336cf9850781d9a8a829fff392dc1
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/20230918/166a87d4/attachment-0001.html>
More information about the ghc-commits
mailing list