Unfortunately I do not have a test case for this (although I can link
[this package
test](60f1be9037/tests/test_simple_combined.nim)
which broke), but this is a regression caused by #24841 (which was
backported to 2.2.4) that causes the following compiler crash:
```
assertions.nim(34) raiseAssert
Error: unhandled exception: ccgtypes.nim(230, 13) `false` mapType: tyGenericInvocation [AssertionDefect]
```
Codegen is traversing the type of the symbol of an explicit destructor
call, but the symbol is the uninstantiated generic hook. This happens
because #24841 changed the code which gives explicit destructor calls
the proper attached destructor to use `replaceHookMagic`, which now
skips `abstractVar` from the type to get the destructor whereas
previously it was just `{tyAlias, tyVar}`. This skips `tyGenericInst`
and also `tyDistinct`. I cannot explain why the skipped `tyGenericInst`
does not have the right destructor but it's not really unexpected, and
skipping `tyDistinct` is just wrong.
To fix this, just `{tyAlias, tyVar, tySink}` are skipped.
(cherry picked from commit 0dc577a4dc)
```nim
type
K = enum
k1,k2
Variant = object
case kind: K
of k1:
discard
of k2:
discard
proc a(x: var K) = discard
proc b(x: ptr K) = discard
var x = Variant(kind: k1)
{.cast(uncheckedAssign).}:
# must be within uncheckedAssign to work
a(x.kind)
b(addr x.kind)
```
(cherry picked from commit 184d423779)
Fixes#25704
This makes sure that `iter` still has `tyGenericInst` skipped like
before, without skipping it for `iterType` which requires it
(cherry picked from commit f9524861f3)
Nested transformBody/liftLambdas passes used a fresh DetectionPass, so
getEnvTypeForOwner could allocate a duplicate PType for the same owner
while :envP already referenced the inner pass type. When addClosureParam
saw cp.typ != t, it errored.
If both types are env objects for the same routine owner, reuse cp.typ
and sync ownerToType.
Adds regression test tests/iter/t21242_nested_closure_in_iter.nim.
(cherry picked from commit 0028ea563c)
fixes#25687
This pull request introduces an optimization for sequence (`seq`)
assignments and copies in the Nim compiler, enabling bulk memory copying
for sequences whose element types are trivially copyable (i.e., no GC
references or destructors). This can significantly improve performance
for such types by avoiding per-element loops.
Key changes:
* Added the `elemSupportsCopyMem` function in
`compiler/liftdestructors.nim` to detect if a sequence's element type is
trivially copyable (no GC refs, no destructors).
* Updated the `fillSeqOp` procedure to use a new `genBulkCopySeq` code
path for eligible element types, generating a call to
`nimCopySeqPayload` for efficient bulk copying. Fallback to the
element-wise loop remains for non-trivial types.
[[1]](diffhunk://#diff-456118dde9a4e21f1b351fd72504d62fc16e9c30354dbb9a3efcb95a29067863R665-R670)
[[2]](diffhunk://#diff-456118dde9a4e21f1b351fd72504d62fc16e9c30354dbb9a3efcb95a29067863R623-R655)
* Introduced the `nimCopySeqPayload` procedure in
`lib/system/seqs_v2.nim`, which performs the actual bulk memory copy of
sequence data using `copyMem`. This is only used for types that are safe
for such an operation.
These changes collectively improve the efficiency of sequence operations
for simple types, while maintaining correctness for complex types.
refc: 3.52s user 0.02s system 99% cpu 3.538 total
orc (after change): 3.46s user 0.01s system 99% cpu 3.476 total
---------
Co-authored-by: Copilot <175728472+Copilot@users.noreply.github.com>
(cherry picked from commit 854c1f15ba)
This fixes type resolution for `iterable[T]`.
I want to proceed with RFC
[#562](https://github.com/nim-lang/RFCs/issues/562) and this is the main
blocker for composability.
Fixes#22098 and, arguably, #19206
```nim
import std/strutils
template collect[T](it: iterable[T]): seq[T] =
block:
var res: seq[T] = @[]
for x in it:
res.add x
res
const text = "a b c d"
let words = text.split.collect()
doAssert words == @[ "a", "b", "c", "d" ]
```
In cases like `strutils.split`, where both proc and iterator overload
exists, the compiler resolves to the `func` overload causing a type
mismatch.
The old mode resolved `text.split` to `seq[string]` before the
surrounding `iterable[T]` requirement was applied, so the argument no
longer matched this template.
It should be noted that, compared to older sequtils templates,
composable chains based on `iterable[T]` require an iterator-producing
expression, e.g. `"foo".items.iterableTmpl()` rather than just
`"foo".iterableTmpl()`. This is actually desirable: it keeps the
iteration boundary explicit and makes iterable-driven templates
intentionally not directly interchangeable with older
untyped/loosely-typed templates like those in `sequtils`, whose internal
iterator setup we have zero control over (e.g. hard-coding adapters like
`items`).
Also, I noticed in `semstmts` that anonymous iterators are always
`closure`, which is not that surprising if you think about it, but still
I added a paragraph to the manual.
Regarding implementation:
From what I gathered, the root cause is that `semOpAux` eagerly
pre-types all arguments with plain flags before overload resolution
begins, so by the time `prepareOperand` processes `split` against the
`iterable[T]`, the wrong overload has already won.
The fix touches a few places:
- `prepareOperand` in `sigmatch.nim`:
When `formal.kind == tyIterable` and the argument was already typed as
something else, it's re-semchecked with the
`efPreferIteratorForIterable` flag. The recheck is limited to direct
calls (`a[0].kind in {nkIdent, nkAccQuoted, nkSym, nkOpenSym}`) to avoid
recursing through `semIndirectOp`/`semOpAux` again.
- `iteratorPreference` field `TCandidate`, checked before
`genericMatches` in `cmpCandidates`, gives the iterator overload a win
without touching the existing iterator heuristic used by `for` loops.
**Limitations:**
The implementation is still flag-driven rather than purely
formal-driven, so the behaviour is a bit too broad `efWantIterable` can
cause iterator results to be wrapped as `tyIterable` in
iterable-admitting contexts, not only when `iterable[T]` match is being
processed.
`iterable[T]` still does not accept closure iterator values such
as`iterator(): T {.closure.}`. It only matches the compiler's internal
`tyIterable`, not arbitrary iterator-typed values.
The existing iterator-preference heuristic is still in place, because
when I tried to remove it, some loosely-related regressions happened. In
particular, ordinary iterator-admitting contexts and iterator chains
still rely on early iterator preference during semchecking, before the
compiler has enough surrounding context to distinguish between
value/iterator producing overloads. Full heuristic removal would require
a broader refactor of dot-chain/intermediate-expression semchecking,
which is just too much for me ATM. This PR narrows only the
tyIterable-specific cases.
**Future work:**
Rework overload resolution to preserve additional information of
matching iterator overloads for calls up to the point where the
iterator-requiring context is established, to avoid re-sem in
`prepareOperand`.
Currently there's no good channel to store that information. Nodes can
get rewritten, TCandidate doesn't live long enough, storing in Context
or some side-table raises the question how to properly key that info.
(cherry picked from commit be29bcd402)
fixes#25682
This pull request introduces a fix to the Nim compiler's assignment code
generation logic to better handle statement list expressions, and adds
regression tests to ensure correct behavior when assigning to object
fields via templates. The changes address a specific bug (#25682)
related to assignments using templates with side effects in static
contexts.
**Compiler code generation improvements:**
* Updated the `genAsgn` procedure in `compiler/vmgen.nim` to properly
handle assignments where the left-hand side is a `nkStmtListExpr`
(statement list expression), ensuring all statements except the last are
executed before the assignment occurs.
**Regression tests for assignment semantics:**
* Added new test blocks in `tests/vm/tvmmisc.nim` to verify that
template-based assignments to object fields work as expected in static
contexts, specifically testing for bug #25682.
(cherry picked from commit 9c07bb94c1)
fixes#25632fixes#25631fixes#25630
This pull request introduces a compatibility check between the
`{.error.}` and `{.exportc.}` pragmas in procedure declarations.
Specifically, it prevents a procedure from being marked with both
pragmas at the same time, as this combination is now considered invalid.
Pragma compatibility enforcement:
* Added a check in `semProcAux` (in `compiler/semstmts.nim`) to emit a
local error if a procedure is declared with both `{.error.}` and
`{.exportc.}` pragmas, preventing their incompatible usage.
(cherry picked from commit fb31e86537)
fixes#25677;
fixes#25678
This pull request introduces both a bug fix to the type checking logic
in the compiler and new test cases for lent types involving procedures
and tables. The most significant change is a refinement in how type
flags are handled for procedure and function types in the compiler,
which improves correctness in type allowance checks. Additionally, the
test suite is expanded to cover more complex scenarios with lent types
and table lookups.
**Compiler improvements:**
* Refined the handling of type flags in `typeAllowedAux` for procedure
and function types by introducing `innerFlags`, which removes certain
flags (`taObjField`, `taTupField`, `taIsOpenArray`) before recursing
into parameter and return types. This ensures more accurate type
checking and prevents inappropriate flag propagation.
**Testing enhancements:**
* Added new test blocks in `tests/lent/tlents.nim` to cover lent
procedure types stored in objects and used as table values, including a
function that retrieves such procedures from a table by key.
* Introduced a test case for an object containing a lent procedure
field, ensuring correct behavior when accessing and using these fields.
(cherry picked from commit 7a82c5920c)
Fixes bug #25670.
The second argument to `max` in `cmpDecimalsIgnoreCase` used `limitB -
iA` instead of `limitB - iB`, which could mis-order numeric segments
when sorting doc index entries.
(cherry picked from commit 5c86c1eda9)
Fixes bug #25671.
The previous condition `not value > 0` was parsed as `(not value) > 0`,
not `not (value > 0)`, so the check did not reliably enforce a positive
`--maxLoopIterationsvm` limit. Align with `--maxcalldepthvm` by using
`value <= 0`.
(cherry picked from commit 7f6b76b34c)
It seems in dispute whether changes to code induced to avoid this new
warning firing are worthwhile.
Until either the analyzer is better or a palatable way to adjust stdlib
code not warn is found, verbosity=1 should not include the warning.
Possibly higher levels, too, but this PR is conservative and only takes
it out at the 2->1 transition.
(cherry picked from commit 4bf44ca47f)
Follows up #25269, refs #25265.
I hit the same bug as #25265 for my own project but #25269 does not fix
it, I think because the type in my case is a `tyGenericInst` which does
not trigger the generation here. First I thought of skipping abstract
type kinds instead of checking for a raw `tyObject`, which fixes my
problem. But in general this could maybe also be encountered for
`tyTuple` and `tySequence` etc. So I figured it might just be safest to
not filter on specific type kinds, ~~which is done now~~ (edit: broke
CI). Maybe this has a slight cost on codegen performance though.
Edit: Allowing all types failed CI for some reason as commented below,
trying skipped type version again.
(cherry picked from commit 1a1586a5fb)
This fixes autogenerated references within the same-module for types,
variables and constants for custom output file names. Previously, the
module name was baked-in, now intra-module links omit the page name in
href.
In short, fixes symbol anchors for `-o:index.html`
Expected test results updated.
(cherry picked from commit 2db13e05ac)
`getTypeImpl` and friends were always putting `nkEmpty` in the default
value field which meant the default values couldn't be introspected.
This copies the default AST so it can be seen in the returned object
(cherry picked from commit edbb32e4c4)
On simple code like:
```nim
type Foo = object
case x: range[0..7]
of 0..2:
a: string
else:
b: string
var foo = Foo()
{.cast(uncheckedAssign).}:
foo.x = 5
```
The compiler tries to generate a destructor for the variant fields by
checking if the discrim is equal to the old one, but the type is not
skipped when looking for an `==` operator in system, so any
discriminator with type `range`/`distinct`/etc crashes with:
```
(10, 9) Error: can't find magic equals operator for type kind tyRange
```
This is fixed by just skipping abstract types.
(cherry picked from commit 7a87e7d199)
fixes#25262
```nim
if constraint != nil and constraint.kind == tyTypeDesc:
n[i].typ = e.typ
else:
n[i].typ = e.typ.skipTypes({tyTypeDesc})
```
at least when `constraint` is a typedesc, it should not skip
`tyTypeDesc`
```nim
if arg.kind != tyTypeDesc:
arg = makeTypeDesc(m.c, arg)
```
Wrappers literals into typedesc, which can cause problems. Though, it
doesn't seem to be necessary
(cherry picked from commit bd709f9b4c)
fixes#25005
In `semTypeIdent`, when resolving a typedesc parameter inside a generic
instantiation, the code took a shortcut: it returned the symbol of the
element type (`bound = result.typ.elementType.sym`). However, for
generic types like `RpcResponse[T] = ref object`, the instantiated
object type (e.g., `RpcResponse:ObjectType[string]`) is a copy with a
new type ID but still points to the same symbol as the uninstantiated
generic body type. That symbol's .typ refers to the original
uninstantiated type, which still contains unresolved generic params `T`
(cherry picked from commit e58acc2e1e)
fixes#25475
```nim
var x: array[0..1, int] = [0, 1]
var y: array[4'u..5'u, int] = [0, 3]
echo x == y
```
sigmatch treats array compatibility by element type + length, not by the
index (range) type. Perhaps backend should do the same check
(cherry picked from commit 97fed258ed)
This fixes two issues with impotc'ed types.
1. Passing an importc'ed inherited object to where superclass is
expected emitted `v.Sup` previously. Now it emits `v`, similar to cpp
codegen.
2. Casting between different nim types that resolve to the same C type
previously was done like `*(T*)&v`, now it is just `v`.
(cherry picked from commit 937e647f4f)
`hashType` proc returned the same hash value from different instanced
generics types like `D[int64]` and `D[F]`.
That caused the struct type with wrong field types.
object/tuple type size check code is generated when it is compiled with
`-d:checkAbi` option.
(cherry picked from commit 88e7adfcb7)
Implements #21928
Adds a `--raw` (since thats what the original issue used, suggestions
welcome) switch which stops the jsondoc gen from rendering rst/markdown.
Implemented by making `genComment` check if it needs to return the raw
string or not. This required switching the related procs to using
`Option` to handle how `nil` values were returned before. The `nil`
returns were eventually ignored so just ignoring `none(T)` has the same
effect.
Doesn't support `runnableExamples` since jsondocs doesn't support them
either
(cherry picked from commit c1e381ae8d)
