GitLab

Duncan Coutts pushed to branch wip/dcoutts/io-manager-tidy at Glasgow Haskell Compiler / GHC

Commits:

66d1a5d5

by fendor at 2026-07-07T16:57:56-04:00

Add 'backendInfoTableMapValidity' backend predicate

Check whether the backend supports the `-finfo-table-map` flag and
ignore it otherwise.

Improve by-design documentation of `backendCodeOutput`.

`Backend` is **abstract by design**. Make this clearer in
`backendCodeOutput` which is incorrectly being used as a proxy for
`Backend`.

Instead, define the desired property predicates in GHC.Driver.Backend
In the process, make `backendCodeOutput` total.

74f1071d

by fendor at 2026-07-07T16:57:56-04:00

Add failing test for `-finfo-table-map` and bytecode backend

If you compile a module using the bytecode backend, with
-finfo-table-map, then the info table map doesn't get populated for the
module.

This is because the -finfo-table-map code path is implemented mostly in
the StgToCmm phase which isn't run when creating bytecode.

Ticket #27039

28d63bca

by mangoiv at 2026-07-07T16:59:16-04:00

ci: don't fail nightly if there have been no changes that night

Fixes #27127

4ebfc478

by Rodrigo Mesquita at 2026-07-08T04:47:53-04:00

ttg: Using ShortText over FastString in the AST

To make the AST independent of GHC, this commit replaces usages of
`FastString` with `HText` in the AST, killing the last edge from
Language.Haskell.* to GHC.* modules.

Even though we /do/ want to use FastStrings in general -- critically in
Names or Ids -- there is no particular reason for the FastStrings that
occur in the AST proper to be FastStrings. Strings in the AST are
typically unique and don't benefit particularly from being interned
FastStrings with Uniques for fast comparison.

`HText` is a type synonym for `ShortText` which uses GHC's Modified
UTF-8 encoding exclusively.

Modified UTF-8 must be used to represent the Haskell AST because the
Haskell Report allows surrogate code points. `Data.Text.Text` functions
use Standard UTF-8 which replace surrogates with a placeholder value,
thus `Data.Text.Text` is unsuitable for AST strings. See the
`Language.Haskell.Syntax.Text` module header for more details.

Final progress towards #21592

Closes #21628

d910b353

by Simon Peyton Jones at 2026-07-08T04:48:36-04:00

Update equality-type documenation in GHC.Builtin.Types.Prim

Fix #27466

b2530542

by Simon Peyton Jones at 2026-07-08T04:48:36-04:00

Honour -dsuppress-coercions in GHC.Core.TyCo.pprCo

Fixes #27467

9a73179a

by Facundo Domínguez at 2026-07-08T04:49:26-04:00

Add item to MR checklist asking to squash fixup commits after approval

The checklist has an item that reads

    All commits are either individually buildable or squashed.

This item could be checked immediately after sending the merge request
though. If reviewers ask for amends later on, and the author amends
the merge request, there was no item that would remind the contributors
to squash the fixup commits before landing.

This commit adds a new item

    After all approvals and before landing: all fixup commits are squashed with their originating commits.

which should be harder to mark as done before approvals have been given.

ed09895d

by Andreas Klebinger at 2026-07-08T16:53:27-04:00

Fix a profiling race condition resulting in segfaults.

StgToCmm: Don't assume tagged FUN closures in closureCodeBody.
When entering a closure the self/node pointer might not be tagged in
some situations when a thunk is evaluated by multiple threads.
So we most AND away the tag bits rather than subtracting an expected tag.

Apply.cmm: Fix a race condition occuring when a thunk is mutated during GC.

In stg_ap_0_fast when might need to run GC before entering a thunk. If this happens
another thread or the GC itself might mutate the closure making entering it no longer
valid. We now check for this.

Add test and changelog for #27123 fixes.

67c03eb2

by Cheng Shao at 2026-07-08T16:54:09-04:00

ghc-heap: fix invalid srtlen returned by peekItbl when no-TNTC

This patch fixes the no-TNTC code path of `peekItbl` so that it looks
at the right memory address when reading the `srt` field from the
`StgInfoTable_` struct. Also adds a `T27465` regression test that
reproduces the bug on no-TNTC builds before the fix. Fixes #27465.

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

2ecabb4f

by Zubin Duggal at 2026-07-09T09:23:25-04:00

hadrian: binary-dist-dir should not be the default target

Revert behaviour to pre 23c9b6c392f52ec9d7a8618b204ff6b885f5fba2

In 23c9b6c392f52ec9d7a8618b204ff6b885f5fba2, we applied the following behaviour change:

```
hadrian: Build stage 2 cross compilers
...
* 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
```

This is a major regression to development experience, a plain hadrian/build
--freeze1 now takes ages because we rebuild all docs (which need to go in the
binary dist dir).

`binary-dist-dir` is the wrong default target for regular GHC development work

Fixes #27445

e16388e3

by Zubin Duggal at 2026-07-09T09:23:25-04:00

.gitignore: Add the hadrian system.config introduced by commit 23c9b6c392f52ec9d7a8618b204ff6b885f5fba2

Since

commit 23c9b6c392f52ec9d7a8618b204ff6b885f5fba2
Author: Matthew Pickering <matthewtpickering@gmail.com>
Date:   Thu Dec 21 16:17:41 2023 +0000

    hadrian: Build stage 2 cross compilers

./configure produces /hadrian/cfg/system.config.{host,target}

Add these to .gitignore

7e8abf41

by Alan Zimmerman at 2026-07-09T09:24:12-04:00

EPA: Replace AnnListItem with simply [TrailingAnn]

Remove the unnecessary wrapper around a single field.

29032f17

by Zubin Duggal at 2026-07-09T09:24:58-04:00

testsuite: Keep real reason for fragile test failures

c34e03a7

by Zubin Duggal at 2026-07-09T09:24:58-04:00

testsuite: Fall back to the failure reason for empty JUnit bodies

409d40f0

by Zubin Duggal at 2026-07-09T09:24:58-04:00

testsuite: Show output diffs in JUnit output

Also refactor compare_outputs to return essentially a `Maybe Diff`
(`CompareOutput`) instead of a bool, but more pythonic. This
allows us to pass the diff through nice.

06fee1ab

by Zubin Duggal at 2026-07-09T09:24:58-04:00

perf notes: include stat deviation and acceptance window in notes so they show up in gitlab

57c0f32c

by mangoiv at 2026-07-10T11:08:38-04:00

driver: enable -finter-module-far-jumps by default

this fixes a compatibility bug with certain binutils/gcc versions where
we were seeing jump offset overflow errors.
This commit can probably reverted if we stop supporting the problematic
binutils/gcc verions (2.44 and 14.2, respectively)

Reolves #26994

4396a6f2
by Andrea Vezzosi at 2026-07-10T11:09:25-04:00
```
[Fix #27287] preserve ModBreaks in ModIface
```

ed261a7e

by Cheng Shao at 2026-07-14T17:59:38-04:00

hadrian: fix HLS support

This patch fixes hadrian's HLS support so one can rely on HLS when
working on the hadrian codebase. Fixes #27480.

Not building/linking shared libraries for hadrian is a severely
premature optimization; this top-level setting in `cabal.project` only
affects home packages while the dependencies in the cabal store are
built with vanilla/dynamic anyway, and even adding dynamic builds to
home packages would not be costly due to cabal's usage of
`-dynamic-too`.

eee8ec5b

by Cheng Shao at 2026-07-14T18:00:20-04:00

compiler: fix miscompiled %load_relaxed, add missing %store_relaxed

This patch fixes the %load_relaxed cmm primop compilation logic to
correctly use relaxed memory ordering, and adds the missing
%store_relaxed primop. Parsing logic of %load/%store with explicit
ordering is covered in the AtomicFetch test case. Fixes #27483.

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

1718230f

by Alan Zimmerman at 2026-07-14T18:01:06-04:00

EPA: Keep binds and sigs together in HsValBindsLR

We combine them into a single list for GhcPs, wrapped in the
ValBind data type, which is the bind equivalent of ValD, having
constructors for binds and sigs.

This simplifies exact print processing, especially when using it to
update the contents of local binds, as we no longer need AnnSortKey
BindTag

6bd1ad2a

by Andreas Klebinger at 2026-07-14T18:01:49-04:00

Bump nofib submodule to account for MonoLocalBinds.

New versions of GHC enable MonoLocalBinds by default.
This breaks some of the benchmarks. I've fixed this and
this bump pulls in that fix.

7eb0f1c9

by Cheng Shao at 2026-07-14T18:02:31-04:00

testsuite: fix bytecodeIPE test under +ipe flavours

This patch fixes the bytecodeIPE test under +ipe flavours. It used to
fail under +ipe because the RTS is built with IPE info, then
stg_AP_info in RTS carries IPE info, so whereFrom wouldn't return
Nothing. Now the test checks IPE info of a datacon in the ghci-loaded
module which is not affected by whether the RTS is built with IPE info
or not. Fixes #27498.

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

ef038aae

by cydparser at 2026-07-15T04:35:41-04:00

Reduce bytes allocated for `capabilities` in RTS (fixes #27487)

In rts/Capability.c, `capabilities` is an array of pointers, but it was allocated as if it were an
array of Capability's.

d377e83e

by Cheng Shao at 2026-07-15T04:36:27-04:00

rts: fix missing UNTAG in stg_readTVarIOzh

This patch fixes missing UNTAG on the current value closure read from
StgTVar. UNTAG is a no-op when it's stg_TREC_HEADER_info which is word
aligned; it may be a tagged closure, and reading info table from the
tagged address is an unaligned load which may cause issues on
platforms with strict alignment requirements.

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

8ed03842

by Cheng Shao at 2026-07-15T04:36:27-04:00

rts: fix missing UNTAG in stg_control0zh_ll

This patch fixes missing UNTAG on the cont closure returned by
captureContinuationAndAbort. In case it's not NULL,
captureContinuationAndAbort returns a tagged StgContinuation closure,
in which case it must be untagged before accessing the
apply_mask_frame field.

In the past it worked out of luck: when apply_mask_frame was NULL then
mask_frame_offset is also 0 so the control flow didn't diverge to a
wrong path. Still, this is horribly wrong and will crash once
StgContinuation struct is refactored and fields are shuffled around.

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

5aa7000a

by Cheng Shao at 2026-07-15T04:37:08-04:00

compiler: fix redundant AP thunk codegen when not using -ticky-ap-thunk

This patch fixes a double negation confusion in !7525 that results in
some redundant AP thunk code generation when not using
-ticky-ap-thunk. Now, we use `stgToCmmUseStdApThunk` to indicate
whether precomputed AP thunks in the RTS should be used, which
defaults to `True`, unless `-ticky-ap-thunk` is passed.
`-finfo-table-map` now also implies `-ticky-ap-thunk`, since when
doing IPE profiling we want the generated AP thunks to be unique.
Fixes #27502.

-------------------------
Metric Decrease:
    T3064
-------------------------

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

d43a7b7a

by Brian McKenna at 2026-07-15T20:10:04+02:00

Strip ticks when desugaring bool guards

The special `considerAccessible` pattern was broken when compiling
with debug info. Compiling with debug info wraps expressions with
`SourceNote` ticks, which broke the internals of the
`desugarBoolGuard` function. Ticks are now ignored within this
function.

Fixes #27360

ede4b17b

by Ben Gamari at 2026-07-15T22:59:53-04:00

base: Display ExceptionContext in WhileHandling's textual description

As originally-implemented the implementation for
`WhileHandling(displayExceptionAnnotation)` would display the
`ExceptionContext` of the exception which it carries (as this was the
behavior of `displayException`, in terms of which
`displayExceptionAnnotation` was implemented).

However, in 284ffab3 the definition of `SomeException(displayException)`
was changed to exclude the `ExceptionContext`. This means that
`WhileHandling(displayExceptionAnnotation)` fails to describe the
provenance of the exception which it captures, greatly limiting its
utility.

Return the implementation to its originally-specified behavior by
implementing `WhileHandling(displayExceptionAnnotation)` in terms of
`displayExceptionWithInfo`.

Fixes #27456.

4f54bcc1

by Duncan Coutts at 2026-07-16T09:14:38+01:00

Move THREADED_RTS-conditional struct members to end of Capability

Accessing members of the Capability struct from CMM code rely on
accessor macros. (The macros are generated by deriveConstants).

These macros have a single definition. This means that the offsets of
all struct members must *not* vary based on THREADED_RTS vs
!THREADED_RTS. This requires that any struct members that are
conditional on THREADED_RTS must occur after the unconditional struct
members. Hence we move all the ones that are conditional on
THREADED_RTS to the end.

Add a deriveConstants entry for the iomgr member of the Capability
struct, which was the motivation for this change.

Add warning messages to help our future selves. Debugging this took me
a couple hours in gdb!

74adf094

by Duncan Coutts at 2026-07-16T09:14:38+01:00

Make the IOManager API use CapIOManager rather than Capability

This makes the API somewhat more self-contained and more consistent.
Now the IOManager API and each of the backends takes just the I/O
manager structure. Previously we had a bit of a mixture, depending on
whether the function needed access to the Capability or just the
CapIOManager.

We still need access to the cap, so we introduce a back reference to
reach the capability, via iomgr->cap.

Convert all uses in select and poll backends, but not win32 ones.
Convert callers in the scheduler and elsewhere.

Also convert the three CMM primops that call IOManager APIs. They just
need to use Capability_iomgr(MyCapability()).

7ef31daf

by Duncan Coutts at 2026-07-16T09:14:38+01:00

Split posix/MIO.c out of posix/Signals.c

The MIO I/O manager was secretly living inside the Signals file.

Now it gets its own file, like any other self-respecting I/O manager.

12382f85

by Duncan Coutts at 2026-07-16T09:14:38+01:00

Rationalise some scheduler run queue utilities

Move them all to the same place in the file.

Make some static that were used only internally.

Also remove a redundant assignment after calling truncateRunQueue that
is already done within truncateRunQueue.

87367a1d

by Duncan Coutts at 2026-07-16T09:14:38+01:00

Rename initIOManager{AfterFork} to {re}startIOManager

These are more accurate names, since these actions happen after
initialisation and are really about starting (or restarting) background
threads.

e7326bb0

by Duncan Coutts at 2026-07-16T09:14:38+01:00

Free per-cap I/O managers during shutdown and forkProcess

Historically this was not strictly necessary. The select and win32
legacy I/O managers did not maintain any dynamically allocated
resources. The new poll one does (an auxillary table), and so this
should be freed.

After forkProcess, all threads get deleted. This includes threads
waiting on I/O or timers. So as of this patch, resetting the I/O
manager is just about tidying things up. For example, for the poll
I/O manager this will reset the size of the AIOP table (which
otherwise grows but never shrinks).

In future however the re-initialising will become neeecessary for
functionality, since some I/O managers will need to re-initialise
wakeup fds that are set CLOEXEC.

76125f45

by Duncan Coutts at 2026-07-16T09:14:38+01:00

Add a TODO to the MIO I/O manager

The direction of travel is to make I/O managers per-capability and have
all their state live in the struct CapIOManager. The MIO I/O manager
however still has a number of global variables.

It's not obvious how handle these globals however.

73ddbe67
by Duncan Coutts at 2026-07-16T09:14:38+01:00
```
Add a FIXME note in the Poll I/O manager
```

3feb9efe