• monkyyy (23/23) Oct 12 Alias reassignment has various bugs, including regressions, in an
• Richard (Rikki) Andrew Cattermole (7/34) Oct 12 If we were to do something like this, it would be a pragma as a hint

monkyyy <crazymonkyyy gmail.com> writes:

Alias reassignment has various bugs, including regressions, in an 
attempt to optimize it. Sometimes I want a symbol to survive, 
other time I(by which i mean core devs) want it deleted and 
hoping a redesign to a static foreach improves compile times.

I expect it will be years before its nailed down and correct.

I suggest two new traits `__traits(delete_alias)` and 
`__traits(keep_alias)`, that explicitly keep or delete aliases a 
work arounds

```d
alias bar=Seq!();
static foreach(...){
   ...//something that triggers bugs
}
bar=__traits(delete_alias);//marks bar for deletion
```

```d
alias bar=Seq!();
bar=__traits(keep_alias);//turn off optimizations
static foreach(...){
   ...//something that triggers bugs
}
foo=bar;//foo should be correct now
```

"Richard (Rikki) Andrew Cattermole" <richard cattermole.co.nz> writes:

On 13/10/2024 7:47 AM, monkyyy wrote:
 Alias reassignment has various bugs, including regressions, in an 
 attempt to optimize it. Sometimes I want a symbol to survive, other time 
 I(by which i mean core devs) want it deleted and hoping a redesign to a 
 static foreach improves compile times.
 
 I expect it will be years before its nailed down and correct.
 
 I suggest two new traits `__traits(delete_alias)` and 
 `__traits(keep_alias)`, that explicitly keep or delete aliases a work 
 arounds
 
 ```d
 alias bar=Seq!();
 static foreach(...){
    ...//something that triggers bugs
 }
 bar=__traits(delete_alias);//marks bar for deletion
 ```
 
 ```d
 alias bar=Seq!();
 bar=__traits(keep_alias);//turn off optimizations
 static foreach(...){
    ...//something that triggers bugs
 }
 foo=bar;//foo should be correct now
 ```

If we were to do something like this, it would be a pragma as a hint 
instead.

That doesn't need a DIP as it is a compiler specific instruction.

Needs a bug report, with behavior of what you want done.

Given its tuning related of existing behavior, you may have some success 
doing a PR for it.