View Issue Details
| ID | Project | Category | View Status | Date Submitted | Last Update |
|---|---|---|---|---|---|
| 0037085 | FPC | Compiler | public | 2020-05-16 11:48 | 2020-05-27 23:26 |
| Reporter | Thaddy de Koning | Assigned To | Florian | ||
| Priority | normal | Severity | minor | Reproducibility | always |
| Status | resolved | Resolution | fixed | ||
| Platform | all | OS | all | ||
| Product Version | 3.3.1 | ||||
| Fixed in Version | 3.3.1 | ||||
| Summary | 0037085: In ISO mode, Dispose with multiple parameters is not yet implemented. | ||||
| Description | According to ISO, if new() is used with parameters, dispose should also use parameters. There is an example on the forum here: https://forum.lazarus.freepascal.org/index.php/topic,49834.msg362392.html#msg362392 Why did this not show up before? Dunno. | ||||
| Tags | No tags attached. | ||||
| Fixed in Revision | 45519 | ||||
| FPCOldBugId | |||||
| FPCTarget | - | ||||
| Attached Files |
| ||||
|
|
Mind, multiple parameters (array of const style) and New() and Dispose() should match when used? It has implications for memory layout. Frankly, I did not know it existed and - apart from the forum example - never saw it used. I think this is hard to fix, maybe just document it as incompatible for now. (Scott, do you have more info?) |
|
|
It would also be interesting to hear if compilers really implement the partial allocation, or if it is an artefact. |
|
|
FPC at least does not implement partial allocation. The only thing it does if a variant selector is provided is to assign the provided value if it is a named variant selector. So in the example provided on the forum there won't be any difference to not using a variant selector. In the following example however the field b will have the value True. {$mode iso}
type
v = ^x;
x = record
n: Integer;
case b: Boolean OF
True: (x0: Real);
False: (x1, x2: Integer)
end;
var
a: v;
begin
New(a, True);
end.That said FPC definitely needs to parse this variant Dispose as well. |
|
|
It is parsed and semantically checked. The generated code does not change though. If it is common, that an RTE is generated if the record contents and the constants do not match, I can implement this as well. |
| Date Modified | Username | Field | Change |
|---|---|---|---|
| 2020-05-16 11:48 | Thaddy de Koning | New Issue | |
| 2020-05-16 12:26 | Thaddy de Koning | Note Added: 0122850 | |
| 2020-05-16 12:27 | Thaddy de Koning | Note Edited: 0122850 | View Revisions |
| 2020-05-16 12:28 | Thaddy de Koning | Note Edited: 0122850 | View Revisions |
| 2020-05-16 13:40 | Marco van de Voort | Note Added: 0122853 | |
| 2020-05-17 16:04 | Sven Barth | Note Added: 0122883 | |
| 2020-05-27 23:25 | Florian | Assigned To | => Florian |
| 2020-05-27 23:25 | Florian | Status | new => resolved |
| 2020-05-27 23:25 | Florian | Resolution | open => fixed |
| 2020-05-27 23:25 | Florian | Fixed in Revision | => 45519 |
| 2020-05-27 23:25 | Florian | FPCTarget | => - |
| 2020-05-27 23:25 | Florian | Note Added: 0123102 | |
| 2020-05-27 23:26 | Florian | Fixed in Version | => 3.3.1 |
