HashMapper needing .dup for stack allocation workaround #164

New Issue

deavmi · 2024-02-27T20:25:35Z

deavmi commented

2024-02-27 20:25:35 +00:00

HashMapper

For some reason the value returned is some stack allocation and hence we need to dup it to get a heap-allocated copy we can actually return. Else we end up returning a stack pointer which then gets overwritten before usage.

Code below:

version(unittest) { writeln(format("hashMapper, prior to hashing the symbol name is: '%s'", path)); }

// FIXME: Needs dup, somewhere this is allocating KAK!
// Some bizarre stack-allocation (maybe) and then returning that in [len, stackPtr]?
string mappedSymbol = toHexString!(LetterCase.lower)(md5Of(path)).dup;

version(unittest) { writeln(format("hashMapper, AFTER hashing symbol name is: '%s'", mappedSymbol)); }

## HashMapper For some reason the value returned is some stack allocation and hence we need to `dup` it to get a heap-allocated copy we can actually return. Else we end up returning a stack pointer which then gets overwritten before usage. Code below: ```d version(unittest) { writeln(format("hashMapper, prior to hashing the symbol name is: '%s'", path)); } // FIXME: Needs dup, somewhere this is allocating KAK! // Some bizarre stack-allocation (maybe) and then returning that in [len, stackPtr]? string mappedSymbol = toHexString!(LetterCase.lower)(md5Of(path)).dup; version(unittest) { writeln(format("hashMapper, AFTER hashing symbol name is: '%s'", mappedSymbol)); } ```

deavmi added this to the Clean ups milestone 2024-02-27 20:25:35 +00:00

deavmi added the

qol

label 2024-02-27 20:25:35 +00:00

deavmi self-assigned this 2024-02-27 20:25:35 +00:00

deavmi added this to the Code emit project 2024-02-27 20:25:35 +00:00

deavmi added spent time 2024-02-27 20:57:02 +00:00

20 minutes

deavmi added the

hashmapper

label 2024-02-27 20:58:44 +00:00

deavmi added the due date 2024-04-27

2024-02-27 20:59:05 +00:00

deavmi changed title from ~~Not so lekker things~~ to HahsMapper needing .dup for stack allocation workaround

2024-02-27 20:59:28 +00:00

deavmi commented

2024-02-27 20:59:39 +00:00

I added the .dup but I don't like that solution.

I added the `.dup` but I don't like that solution.

deavmi changed title from ~~HahsMapper needing .dup for stack allocation workaround~~ to HashMapper needing .dup for stack allocation workaround

2024-02-27 20:59:48 +00:00

deavmi removed the due date 2024-04-27

2024-04-01 21:26:35 +01:00

deavmi added the due date 2024-04-05

2024-04-01 21:26:43 +01:00

deavmi commented

2024-04-07 12:18:33 +01:00

I should probably look at this again/

deavmi commented

2024-04-07 12:31:32 +01:00

// We do .dup because the hex string is allocated on the stack and
// returning that wll only be valid still the stack grows into the
// space it uccopies again later, so copy it to the heap rather
// and return that

// We do `.dup` because the hex string is allocated on the stack and // returning that wll only be valid still the stack grows into the // space it uccopies again later, so copy it to the heap rather // and return that

deavmi commented

2024-04-07 12:31:59 +01:00

// We do .dup because the hex string is allocated on the stack and
// returning that wll only be valid still the stack grows into the
// space it uccopies again later, so copy it to the heap rather
// and return that

Removed the .dup because when we use the format string a new one is allocated so it is no lonmger needed:

string mappedSymbol = format("t_%s", toHexString!(LetterCase.lower)(md5Of(path)));

> // We do `.dup` because the hex string is allocated on the stack and > // returning that wll only be valid still the stack grows into the > // space it uccopies again later, so copy it to the heap rather > // and return that Removed the `.dup` because when we use the format string a new one is allocated so it is no lonmger needed: ```d string mappedSymbol = format("t_%s", toHexString!(LetterCase.lower)(md5Of(path))); ```

deavmi added spent time 2024-04-07 12:33:19 +01:00

10 minutes

deavmi started working 2024-04-07 12:33:22 +01:00

deavmi commented

2024-04-07 12:33:38 +01:00

Waiting on semantic tests to run...