Add support for inout lifetime dependence #80452
Conversation
|
@swift-ci test |
…atibleOwnership to inferLifetimeDependenceKind
|
@swift-ci test |
|
@swift-ci test |
|
@swift-ci test |
meg-gupta
requested review from
xymus,
hborla,
slavapestov,
xedin,
eeckstein,
a team,
ahoppen,
bnbarham,
CodaFi,
DougGregor,
hamishknight and
rintaro
as code owners
meg-gupta
removed request for
a team,
xedin,
slavapestov,
rintaro,
xymus,
DougGregor,
bnbarham and
CodaFi
meg-gupta
requested review from
atrick and
glessard
and removed request for
ahoppen,
hborla,
hamishknight and
eeckstein
| @@ -631,6 +631,7 @@ function(_compile_swift_files | |||
|
|
|||
| list(APPEND swift_flags "-enable-experimental-feature" "NonescapableTypes") | |||
| list(APPEND swift_flags "-enable-experimental-feature" "LifetimeDependence") | |||
| list(APPEND swift_flags "-enable-experimental-feature" "InoutLifetimeDependence") | |||
There was a problem hiding this comment.
Why do we need to keep adding experimental feature flags to the stdlib command line? Can't this be a non-experimental "baseline" feature?
There was a problem hiding this comment.
We need the experimental feature to ensure old compilers that recognize LifetimeDependence feature but don't have support for @lifetime(&arg) to parse the new stdlib interface containing this new syntax.
There was a problem hiding this comment.
I wonder if we should support the syntax as a baseline feature but only make it an error to use borrow for an inout argument under the experimental flag?
There was a problem hiding this comment.
Regarding the feature flag, I understand now that it is only meant to protect against an older compiler parsing the new stdlib interface. Any code outside the stdlib that uses MutableSpan simply needs to be updated--we don't expect that to be a problem in practice.
| @@ -8171,6 +8171,9 @@ ERROR(lifetime_dependence_invalid_inherit_escapable_type, none, | |||
| ERROR(lifetime_dependence_cannot_use_parsed_borrow_consuming, none, | |||
| "invalid use of borrow dependence with consuming ownership", | |||
| ()) | |||
| ERROR(lifetime_dependence_cannot_use_default_escapable_consuming, none, | |||
| "invalid lifetime dependence on Escapable value with %0 ownership", | |||
There was a problem hiding this comment.
to be consistent with using full grammar:
dependence on an Escapable value
lib/AST/LifetimeDependence.cpp
Outdated
| diagnose(loc, diag::lifetime_dependence_invalid_inherit_escapable_type, | ||
| descriptor.getString()); | ||
| return std::nullopt; | ||
| } | ||
| return kind; | ||
| return parsedLifetimeKind == ParsedLifetimeDependenceKind::Inherit |
There was a problem hiding this comment.
This could be more self-documenting as a switch over ParsedLifetimeDependenceKind instead of code that does not mention Inout, which is really they thing being handled specially.
| @@ -631,6 +631,7 @@ function(_compile_swift_files | |||
|
|
|||
| list(APPEND swift_flags "-enable-experimental-feature" "NonescapableTypes") | |||
| list(APPEND swift_flags "-enable-experimental-feature" "LifetimeDependence") | |||
| list(APPEND swift_flags "-enable-experimental-feature" "InoutLifetimeDependence") | |||
There was a problem hiding this comment.
I wonder if we should support the syntax as a baseline feature but only make it an error to use borrow for an inout argument under the experimental flag?
|
@swift-ci test |
Add support for spelling lifetime dependencies on mutable sources as
@lifetime(&arg). Previously this was spelled as@lifetime(borrow arg)and caused some confusion to developers. This is purely a syntactic change and does not change how these dependencies are represented in SIL.rdar://146310571