Repository navigation
Conversation
|
Please add a test. |
|
@smalyshev: Tests added now The test for the inifile handler fails because this handler insert items even if the item already exists. I'm not a C developer so it would be the best merging this but leaving the bug open so another person can fix it. |
|
Hello guys, is there someone could review this ? |
|
@marc-mabe sorry for the long delay. Please fix the coding style issues after that it looks good for merge. |
|
@lstrojny : thank you for review! Updated coding style issues now |
There was a problem hiding this comment.
Is this really correct? Shouldn’t be the default case just throw some error?
There was a problem hiding this comment.
The default case should be impossible - it was copied from DBA_UPDATE_FUNC(flatfile)
- What should the message say on this case ?
- "Unknown return value %value% by %lib-function%" ?
- error, warning or notice?
There was a problem hiding this comment.
"Unknown return value" is fine. Warning seems appropriate to me.
|
After fixing the switch case, could you rebase and squash? |
…ected handlers are flatfile, inifile, qdbm and gdbm)
|
done! - updated the switch statements and squashed all commits together |
|
Merged into 5.5 and master. Cannot close it now as GitHub seems to be in trouble. |
|
Comment on behalf of lstrojny at php.net: Merged |
SHARP_MAGO_COMMIT named the head of Mago #125 while it was under review. #125 merged, so the pin now names its merge commit on Mago master. That commit also words an Any comparison operand in PHP#'s words in a .sharp file. The header it generates differs from this one only in the pin. ext/sharp/ └── sharp_unit.h <- SHARP_MAGO_COMMIT pins Mago 449377a, the merge of #125
Problem: the generics slice branched from master c3a4a3c. Master since gained effects on function types (#126), private fields with no modifier (#128), union collections (#116), the sortedBy key rule (#125), macOS and Linux CI (#127), function uses (#122) and message fixes (#117). Ten files conflicted, and one master test named a syntax field the slice renamed. Solution: every conflict keeps both sides' behavior. - A refused expression has the error type in PHP# and `never` in PHP, as master types it, and `is_refused` keeps the slice's wider set of refused generic expressions. - An argument is a possible mismatch only when the parts it may share with the parameter, as `shareable_part` takes them, fit through master's `can_argument_be_identical`, and `verify_argument_type` returns whether the argument fits, as master's effect rules read it. - A method value's signature names the class it is read through, as master needs for `extern` lookups, while the slice still specializes it for its receiver and fixes its type arguments. - A PHP# literal's backed enum key and master's other new calls pass the dialect `get_backing_key_type` takes on the slice. - Comparisons master added use the slice's `ComparisonResult::for_dialect`. crates/ ├── analyzer/src/ │ ├── artifacts.rs tested_types beside function_values │ ├── expression/ │ │ ├── array.rs master's PHP# key branch, with the dialect │ │ ├── assignment/array_assignment.rs every declared element, for_dialect │ │ └── mod.rs slice's is_refused, master's error type │ ├── invocation/arguments.rs shareable_part over can_argument_be_identical │ ├── resolver/ │ │ ├── method.rs declared collections, paired with their written atomic │ │ └── property.rs method value source is the read-through class │ └── statement/ │ ├── class_like/property.rs for_dialect, records fitting function values │ └── function_like/mod.rs passes tested_types and function_values up ├── codex/src/scanner/ttype.rs function type keeps its effects and its type parameters └── syntax/tests/sharp.rs effects test reads type_arguments.arguments
removed warnings if dba_insert failed because of an already existing key.
fixed handlers are: