- Dec 09, 2022
-
-
It is a relic of the Make build system. The RTS now uses a `package.conf` file generated the usual way by Cabal.
-
... thus fixing #22549. The details are in the refurbished and no longer dead `Note [Do not strictify a DFun's parameter dictionaries]`. There's a regression test in T22549.
-
I copied the fusion framework we have in place for `take`. T18964 asserts that we regress neither when fusion fires nor when it doesn't. Fixes #18964.
-
There was some confusion in Data.Typeable about which module certain wired-in things were defined in. Just because something is wired-in doesn't mean it comes from GHC.Prim, in particular things like LiftedRep and RuntimeRep are defined in GHC.Types and that's the end of the story. Things like Int#, Float# etc are defined in GHC.Prim as they have no Haskell definition site at all so we need to generate type representations for them (which live in GHC.Types). Fixes #22510
-
So that we get to cancel away the allocation for the lazily used base. We can move `powImpl` (which *is* strict in the base) to the top-level so that we don't duplicate too much code and move the SPECIALISATION pragmas onto `powImpl`. The net effect of this change is that `(^)` plays along much better with inlining thresholds and loopification (#22227), for example in `x2n1`. Fixes #22324.
-
-
-
-
-
-
In #22217 it was observed that the order modules are compiled in affects the contents of an interface file. This was because a module dependended on another module indirectly, via a re-export but the interface file for this module was never loaded because the symbol was never used in the file. If we decide that we depend on a module then we jolly well ought to record this fact in the interface file! Otherwise it could lead to very subtle recompilation bugs if the dependency is not tracked and the module is updated. Therefore the best thing to do is just to make sure the file is loaded by calling the `loadSysInterface` function. This first checks the caches (like we did before) but then actually goes to find the interface on disk if it wasn't loaded. Fixes #22217
-
There was some code in the signature merging logic which added together the ImportAvails of the signature and the signature which was merged into it. This had the side-effect of making the merged signature depend on the signature (via a normal module dependency). The intention was to propagate orphan instances through the merge but this also messed up recompilation logic because we shouldn't be attempting to load B.hi when mergeing it. The fix is to just combine the part of ImportAvails that we intended to (transitive info, orphan instances and type family instances) rather than the whole thing.
-
The idea is that both the bindists should product libraries with the same ABI and interface hash. So the job checks with ghc-pkg to make sure the computed ABI is the same. In future this job can be extended to check for the other facets of interface determinism. Fixes #22180
-
In this job we can run on every commit we add a test which builds the Cabal library twice and checks that the ABI hash and interface hash is stable across the two builds. * We run the test 20 times to try to weed out any race conditions due to `-j` * We run the builds in different temporary directories to try to weed out anything related to build directory affecting ABI or interface file hash. Fixes #22180
-
-
- Dec 08, 2022
-
-
The Show instance for TypeRep [] has changed in 9.5 to output "List" because the name of the type constructor changed. This seems to be accidental and is inconsistent with TypeReps of saturated lists, which are printed as e.g. "[Int]". For now, I'm restoring the old behavior; in the future, maybe we should show TypeReps without puns (List, Tuple, Type).
-
Hadrian used to pass -g when building all ways of rts. It makes output binaries larger (especially so for wasm backend), and isn't needed by most users out there, so this patch removes that flag. In case the debug info is desired, we still pass -g3 when building the debug way, and there's also the debug_info flavour transformer which ensures -g3 is passed for all rts ways.
-
RTS eventlog events for postCapsetVecEvent are truncated if payload is larger than EVENT_PAYLOAD_SIZE_MAX Previously, postCapsetVecEvent records eventlog event with payload of variable size larger than EVENT_PAYLOAD_SIZE_MAX (2^16) without any validation, resulting in corrupted data. For example, this happens when a Haskell binary is invoked with very long command line arguments exceeding 2^16 bytes (see #20221). Now we check the size of accumulated payload messages incrementally, and truncate the message just before the payload size exceeds EVENT_PAYLOAD_SIZE_MAX. RTS will warn the user with a message showing how many arguments are truncated.
-
This module can be used to construct ill-formed TypeReps, so it should be Unsafe.
-
by building its result connected component by component Fixes #22512
-
-
...another manifestation of #20851 which I unfortunately missed in my first pass.
-
- Dec 06, 2022
-
-
-
-
-
-
Commit 6b2f7ffe changed the logic that decided whether to enable debug assertions. However, it had an off-by-one error, as the stage parameter to the function inconsistently referred to the stage of the compiler being used to build or the stage of the compiler we are building. This patch makes it consistent. Now the parameter always refers to the the compiler which is being built. In particular, this patch re-enables assertions in the stage 2 compiler when building with devel2 flavour, and disables assertions in the stage 2 compiler when building with validate flavour. Some extra performance tests are now run in the "validate" jobs because the stage2 compiler no longer contains assertions. ------------------------- Metric Decrease: CoOpt_Singletons MultiComponentModules MultiComponentModulesRecomp MultiLayerModulesTH_OneShot T11374 T12227 T12234 T13253-spj T13701 T14683 T14697 T15703 T17096 T17516 T18304 T18478 T18923 T5030 T9872b TcPlugin_RewritePerf Metric Increase: MultiComponentModules MultiComponentModulesRecomp MultiLayerModules MultiLayerModulesRecomp MultiLayerModulesTH_Make T13386 T13719 T3294 T9233 T9675 parsing001 -------------------------
-
- Dec 05, 2022
-
-
The key part of this patch is the change to mkTokenLocation: - mkTokenLocation (RealSrcSpan r _) = TokenLoc (EpaSpan r) + mkTokenLocation (RealSrcSpan r mb) = TokenLoc (EpaSpan r mb) mkTokenLocation used to discard the BufSpan, but now it is saved and can be retrieved from LHsToken or LHsUniToken. This is made possible by the following change to EpaLocation: - data EpaLocation = EpaSpan !RealSrcSpan + data EpaLocation = EpaSpan !RealSrcSpan !(Strict.Maybe BufSpan) | ... The end goal is to make use of the BufSpan in Parser/PostProcess/Haddock.
-
- Dec 03, 2022
-
-
Ross Paterson authoredThis adds a TypeDataD constructor to the Template Haskell Dec type, and ensures that the constructors it contains go in the TyCls namespace.
-
Consider this example: [d| instance forall a. C [a] where type forall b. G [a] b = Proxy b |] When we process "forall b." in the associated type instance, it is unambiguously the binding site for "b" and we want a fresh name for it. Therefore, FreshNamesOnly is more fitting than ReuseBoundNames. This should not have any observable effect but it avoids pointless lookups in the MetaEnv.
-
This is a refactoring that should have no effect on observable behavior. Prior to this change, GHC.HsToCore.Quote contained a few closely related functions to process type variable bindings: addSimpleTyVarBinds, addHsTyVarBinds, addQTyVarBinds, and addTyClTyVarBinds. We can classify them by their input type and name generation strategy: Fresh names only Reuse bound names +---------------------+-------------------+ [Name] | addSimpleTyVarBinds | | [LHsTyVarBndr flag GhcRn] | addHsTyVarBinds | | LHsQTyVars GhcRn | addQTyVarBinds | addTyClTyVarBinds | +---------------------+-------------------+ Note how two functions are missing. Because of this omission, there were two places where a LHsQTyVars value was constructed just to be able to pass it to addTyClTyVarBinds: 1. mk_qtvs in addHsOuterFamEqnTyVarBinds -- bad 2. mkHsQTvs in repFamilyDecl -- bad This prevented me from making other changes to LHsQTyVars, so the main goal of this refactoring is to get rid of those workarounds. The most direct solution would be to define the missing functions. But that would lead to a certain amount of code duplication. To avoid code duplication, I factored out the name generation strategy into a function parameter: data FreshOrReuse = FreshNamesOnly | ReuseBoundNames addSimpleTyVarBinds :: FreshOrReuse -> ... addHsTyVarBinds :: FreshOrReuse -> ... addQTyVarBinds :: FreshOrReuse -> ... -
See ghc/ghc#16966
-
Lint was not able to see that x*y <= x*y, because this inequality was decomposed to x <= x*y && y <= x*y, but there was no rule to see that x <= x*y. Fixes #22546.
-
See Note [Root-main Id] The `:Main` special binding is actually defined in the current module (hence don't go looking for it externally) but the module name is rOOT_MAIN rather than the current module so we need this special case. There was already some similar logic in `GHC.Rename.Env` for External Core, but now the "External Core" is in interface files it needs to be moved here instead. Fixes #22405
-
- Dec 02, 2022
-
-
This improves the error messages of the exhaustiveness checker when checking statements which have been moved around with ApplicativeDo. Before: Test.hs:2:3: warning: [GHC-62161] [-Wincomplete-uni-patterns] Pattern match(es) are non-exhaustive In a pattern binding: Patterns of type ‘Maybe ()’ not matched: Nothing | 2 | let x = () | ^^^^^^^^^^ After: Test.hs:4:3: warning: [GHC-62161] [-Wincomplete-uni-patterns] Pattern match(es) are non-exhaustive In a pattern binding: Patterns of type ‘Maybe ()’ not matched: Nothing | 4 | ~(Just res1) <- seq x (pure $ Nothing @()) | Fixes #22483 -
This patch changes the representation of TyCon so that it has a top-level product type, with a field that gives the details (newtype, type family etc), #22458. Not much change in allocation, but execution seems to be a bit faster. Includes a change to the haddock submodule to adjust for API changes.
-
- Dec 01, 2022
-
-
-
These jobs fail (and are allowed to fail) nearly every time. Soon they won't even be able to run at all, as we won't currently have runners that can run them. Fixing the latter problem is tracked in #22409. I went ahead and removed all settings and configurations.
-
To do so, we mark some tests broken in this configuration.
-
This subtle bug showed up when compiling a library with 9.4. See #22491. The bug is present in master, but it is hard to trigger; the new regression test T22491 fails in 9.4. The fix was easy: just add a missing varToCoreExpr in etaBodyForJoinPoint. The fix is definitely right though! I also did some other minor refatoring: * Moved the preInlineUnconditionally test in simplExprF1 to before the call to joinPointBinding_maybe, to avoid fruitless eta-expansion. * Added a boolean from_lam flag to simplNonRecE, to avoid two fruitless tests, and commented it a bit better. These refactorings seem to save 0.1% on compile-time allocation in perf/compiler; with a max saving of 1.4% in T9961 Metric Decrease: T9961
-
