| Commit message (Collapse) | Author | Age | Files | Lines |
|
|
|
|
|
|
|
|
| |
In https://gitlab.haskell.org/ghc/ghc/merge_requests/2192#note_246551
Simon convinced me that ignoring type variables existentially bound by
data constructors have to be the same way as value binders.
Sadly I couldn't think of a regression test, but I'm confident that this
change strictly improves on the status quo.
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
We used to check `GrdVec`s arising from multiple clauses and guards in
isolation. That resulted in a split between `pmCheck` and
`pmCheckGuards`, the implementations of which were similar, but subtly
different in detail. Also the throttling mechanism described in
`Note [Countering exponential blowup]` ultimately got quite complicated
because it had to cater for both checking functions.
This patch realises that pattern match checking doesn't just consider
single guarded RHSs, but that it's always a whole set of clauses, each
of which can have multiple guarded RHSs in turn. We do so by
translating a list of `Match`es to a `GrdTree`:
```haskell
data GrdTree
= Rhs !RhsInfo
| Guard !PmGrd !GrdTree -- captures lef-to-right match semantics
| Sequence !GrdTree !GrdTree -- captures top-to-bottom match semantics
| Empty -- For -XEmptyCase, neutral element of Sequence
```
Then we have a function `checkGrdTree` that matches a given `GrdTree`
against an incoming set of values, represented by `Deltas`:
```haskell
checkGrdTree :: GrdTree -> Deltas -> CheckResult
...
```
Throttling is isolated to the `Sequence` case and becomes as easy as one
would expect: When the union of uncovered values becomes too big, just
return the original incoming `Deltas` instead (which is always a
superset of the union, thus a sound approximation).
The returned `CheckResult` contains two things:
1. The set of values that were not covered by any of the clauses, for
exhaustivity warnings.
2. The `AnnotatedTree` that enriches the syntactic structure of the
input program with divergence and inaccessibility information.
This is `AnnotatedTree`:
```haskell
data AnnotatedTree
= AccessibleRhs !RhsInfo
| InaccessibleRhs !RhsInfo
| MayDiverge !AnnotatedTree
| SequenceAnn !AnnotatedTree !AnnotatedTree
| EmptyAnn
```
Crucially, `MayDiverge` asserts that the tree may force diverging
values, so not all of its wrapped clauses can be redundant.
While the set of uncovered values can be used to generate the missing
equations for warning messages, redundant and proper inaccessible
equations can be extracted from `AnnotatedTree` by
`redundantAndInaccessibleRhss`.
For this to work properly, the interface to the Oracle had to change.
There's only `addPmCts` now, which takes a bag of `PmCt`s. There's a
whole bunch of `PmCt` variants to replace the different oracle functions
from before.
The new `AnnotatedTree` structure allows for more accurate warning
reporting (as evidenced by a number of changes spread throughout GHC's
code base), thus we fix #17465.
Fixes #17646 on the go.
Metric Decrease:
T11822
T9233
PmSeriesS
haddock.compiler
|
| |
|
|
|
|
|
|
|
|
|
|
| |
This patch clarifies a dark corner of quantified
constraints.
* See Note [Yukky eq_sel for a HoleDest] in TcSMonad
* Minor refactor, breaking out new function
TcInteract.doTopReactEqPred
|
| |
|
| |
|
| |
|
|
|
|
| |
The test fails when used with LANG=fr_FR.UTF-8
|
|
|
|
|
| |
Previously we would keep them for two weeks. However, on the stable
branches two weeks can easily elapse with no pushes.
|
| |
|
|
|
|
|
|
|
|
|
| |
This changes GHC's treatment of so-called Naughty Quantification
Candidates to issue errors, instead of zapping to Any.
Close #16775.
No new test cases, because existing ones cover this well.
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
This patch implements overloaded quotation brackets which generalise the
desugaring of all quotation forms in terms of a new minimal interface.
The main change is that a quotation, for example, [e| 5 |], will now
have type `Quote m => m Exp` rather than `Q Exp`. The `Quote` typeclass
contains a single method for generating new names which is used when
desugaring binding structures.
The return type of functions from the `Lift` type class, `lift` and `liftTyped` have
been restricted to `forall m . Quote m => m Exp` rather than returning a
result in a Q monad.
More details about the feature can be read in the GHC proposal.
https://github.com/ghc-proposals/ghc-proposals/blob/master/proposals/0246-overloaded-bracket.rst
|
|
|
|
| |
[skip ci]
|
|
|
|
|
|
|
|
|
|
|
| |
This brings the pretty-printer for Core in line with how visible
type applications are normally printed: namely, with no whitespace
after the `@` character (i.e., `f @a` instead of `f @ a`). While I'm
in town, I also give the same treatment to type abstractions (i.e.,
`\(@a)` instead of `\(@ a)`) and coercion applications (i.e.,
`f @~x` instead of `f @~ x`).
Fixes #17643.
|
|
|
|
| |
(cherry picked from commit feb3b955402d53c3875dd7a9a39f322827e5bd69)
|
| |
|
|
|
|
|
|
|
|
| |
Analyzing the call sites for `HsModule` reveals that it is only ever
used with parsed code (i.e., `GhcPs`). This simplifies `HsModule` by
concretizing its `pass` parameter to always be `GhcPs`.
Fixes #17642.
|
|
|
|
| |
`parallel` is no longer a submodule since 3cb063c805ec841ca33b8371ef8aba9329221b6c
|
|
|
|
|
| |
Otherwise the testsuite driver may not have an up-to-date
baseline.
|
|
|
|
| |
Refactoring to follow.
|
|
|
|
|
|
|
|
| |
Previously we used platform.system() and while this worked fine (e.g.
returned `Windows`, as expected) locally under both msys and MingW64
Python distributions, it inexplicably returned `MINGW64_NT-10.0`
under MingW64 Python on CI. It seems os.name is more reliable so we now
use that instead..
|
|
|
|
|
| |
In addition, we prefer the Mingw64 Python distribution on Windows due
to #17483.
|
| |
|
|
|
|
| |
Due to #17607.
|
| |
|
|
|
|
|
|
|
|
|
|
|
| |
Also refactor FP_GCC_EXTRA_FLAGS in a few ways:
* We no longer support compilers which lack support for -fno-builtin
and -fwrapv so remove the condition on GccVersion
* These flags are only necessary when using the via-C backend
so make them conditional on Unregisterised.
Fixes #15742.
|
| |
|
| |
|
|
|
|
|
|
| |
I have seen this fail both on x86-64/Debian 9 and armv7/Debian 9
See #17554.
|
|
|
|
|
| |
This is to prepare for ghc-bignum which implements some but not all of
gmp functions.
|
|
|
|
|
| |
This a small utility function that comes in handy when debugging the
lexer and the parser.
|
|
|
|
|
| |
Metric Increase:
haddock.Cabal
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
* stgToCmm hook
* cmmToRawCmm hook
These hooks are used by Asterius and could be useful to other clients of
the GHC API.
It increases the Parser dependencies (test CountParserDeps) to 184. It's
still less than 200 which was the initial request (cf
https://mail.haskell.org/pipermail/ghc-devs/2019-September/018122.html)
so I think it's ok to merge this.
|
| |
|
| |
|
|
|
|
|
|
|
|
|
|
| |
When dumping Cmm groups check if the group is empty, to avoid generating
empty sections in dump files like
==================== Output Cmm ====================
[]
Also fixes a few bad indentation in the code around changes.
|
| |
|
| |
|
| |
|
|
|
|
|
|
|
|
|
|
|
|
| |
Fixed tests:
* haddockA039: added to all.T
* haddockE004: replaced with T17561 (marked as expect_broken)
New tests:
* haddockA040: deriving clause for a data instance
* haddockA041: haddock and CPP #include
|
| |
|
|
|
|
|
|
|
|
| |
As noted in #17624, it's quite unstable, especially, for some reason, on
i386 and armv7 (something about 32-bit platforms perhaps?).
Metric Increase:
T1969
|
|
|
|
|
| |
This makes it easier to see the true magnitude of fluctuations.
Also do some house-keeping in the argument parsing department.
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
Previously we would throw away source ticks when the debug level was
non-zero. This is precisely the opposite of what was intended.
Fixes #17616.
Metric Decrease:
T13056
T9020
T9961
T12425
|
|
|
|
| |
Lest we fail to rebuild when they change. Fixes #17611.
|
|
|
|
|
|
| |
LLVM does not guarantee any particular semantics when dereferencing null
pointers. Consequently, this test actually passes when built with the
LLVM backend.
|
|
|
|
|
|
|
|
|
| |
As noted in #17606, Docker disallows the get_mempolicy syscall by
default. This caused numerous tests to fail under CI in the `debug_numa`
way. Avoid this by disabling the NUMA probing logic when --debug-numa is
in use, instead setting n_numa_nodes in RtsFlags.c.
Fixes #17606.
|
|
|
|
|
| |
Previously things like `+RTS --numa-debug` would enable NUMA support,
despite being an invalid flag.
|
|
|
|
|
| |
This was a hack which is no longer necessary now since we introduce a
dedicated entry block for each procedure.
|
|
|
|
|
|
|
|
|
|
| |
The LLVM IR forbids the entry label of a procedure from having any
predecessors. In the case of a simple looping function the LLVM code
generator broke this invariant, as noted in #17589. Fix this by
moving the function prologue to its own basic block, as suggested by
@kavon in #11649.
Fixes #11649 and #17589.
|