On 10/15/2025 10:10 AM, Luna wrote:
> On Wednesday, 15 October 2025 at 05:04:17 UTC, Richard (Rikki) Andrew Cattermole wrote:
>> I'm looking over the OSX build logs atm, and I'm noticing a lot of warnings being generated.
>>
>> These warnings appear to come in the forms:
>>
>> 1. ``warning: could not find referenced DIE`` (debug info)
>>
>> 2. ``warning: (x86_64)  could not find object file symbol for symbol SYMBOL``
>>
>> 3. ``ld: warning: could not create compact unwind for SYMBOL: offset of saved registers too far to encode``
>>
>> See: https://github.com/dlang/dmd/actions/runs/18517485989/job/52770757595?pr=21970
>> https://github.com/dlang/dmd/actions/runs/18517485989/job/52770757597?pr=21970
>>
>> The result is the OSX build for dmd appears not to finish CI and times out at 40m (we should be able to bump that up).
>>
>> Unfortunately, I am of no help to debug let alone fix these issues.
> 
> I plan to track down these issues once the arm64 support lands in DMD, some of the work debugging and fixing osx things in DMD would likely interfere with that work so I don’t want to risk creating a major headache for Walter there.
> 

I haven't worked at all on the AArch64 debug info generation, so I don't think this would be a problem.

On 16/10/2025 9:23 AM, Walter Bright wrote:
>     I plan to track down these issues once the arm64 support lands in
>     DMD, some of the work debugging and fixing osx things in DMD would
>     likely interfere with that work so I don’t want to risk creating a
>     major headache for Walter there.
> 
> I haven't worked at all on the AArch64 debug info generation, so I don't think this would be a problem.

There have been some changes to DWARF in ~master so the suspicion by me and Iain is those changes introduced the regression.

Luna you are good to go to work on this if you want to.

On 10/15/2025 2:25 PM, Richard (Rikki) Andrew Cattermole wrote:
> There have been some changes to DWARF in ~master so the suspicion by me and Iain is those changes introduced the regression.

Hmm. Linky?


> Luna you are good to go to work on this if you want to.

yes please!

On 17/10/2025 7:34 AM, Walter Bright wrote:
> On 10/15/2025 2:25 PM, Richard (Rikki) Andrew Cattermole wrote:
>> There have been some changes to DWARF in ~master so the suspicion by me and Iain is those changes introduced the regression.
> 
> Hmm. Linky?

https://github.com/dlang/dmd/pull/21664

Hmm I was thinking of: https://github.com/dlang/dmd/pull/21912

But that is CodeView not DWARF.

Guess they haven't changed like I thought they did.

Given that I just got an email stating that osx-13 runners is going away and to upgrade to osx-15-intel its probably a good idea to upgrade and see if that solves it.

We will be skipping osx-14 for the runner as it only comes in -large variant and that's paid I think.

Forwarding from Iain that got posted on my (unrelated) PR:

Managed to get round to drilling down a bit, issue is shared structs, so may not be a regression afterall.

```d
struct AllocatorList
{
    void[] allocators;
    void allocate() { } // b
    // p this
    // $1 = (test.AllocatorList &) @0x7fffffffd4d0: {allocators = 0x0}
}

shared struct SharedAllocatorList
{
    void[] allocators;
    void allocate() { } // b
    // p this
    // $2 = (<unknown type in /build/dlang/test, CU 0x0, DIE 0x178> &) @0x7fffffffd4e0:
    // <unknown type in /build/dlang/test, CU 0x0, DIE 0x178>
}

void main()
{
    AllocatorList a;
    a.allocate();
    SharedAllocatorList b;
    b.allocate();
}
```

Iain was able to locate the commit (and me the PR), that introduced shared into DWARF:

https://github.com/dlang/dmd/pull/11568

On 15/10/2025 6:04 PM, Richard (Rikki) Andrew Cattermole wrote:
> I'm looking over the OSX build logs atm, and I'm noticing a lot of warnings being generated.
> 
> These warnings appear to come in the forms:
> 
> 1. ``warning: could not find referenced DIE`` (debug info)
> 
> 2. ``warning: (x86_64)  could not find object file symbol for symbol SYMBOL``
> 
> 3. ``ld: warning: could not create compact unwind for SYMBOL: offset of saved registers too far to encode``
> 
> See: https://github.com/dlang/dmd/actions/runs/18517485989/ job/52770757595?pr=21970
> https://github.com/dlang/dmd/actions/runs/18517485989/job/52770757597? pr=21970
> 
> The result is the OSX build for dmd appears not to finish CI and times out at 40m (we should be able to bump that up).
> 
> Unfortunately, I am of no help to debug let alone fix these issues.

Uck bumping to osx-15 won't be happening.

https://github.com/dlang/dmd/actions/runs/18578513719/job/52968627848?pr=21985

The x87 real's are not turned on (perhaps 15 doesn't have it).

The backend then trys to use the longdouble module, only it is disabled there.

I'll also have to disable the bootstrap CI instance for 14:

ld: multiple errors: symbol count from symbol table and dynamic symbol table differ in '/Users/runner/work/dmd/dmd/generated/build.o' in '/Users/runner/work/dmd/dmd/generated/build.o'; address=0x0 points to section(2) with no content in '/Users/runner/dlang/dmd-2.099.1/osx/lib/libphobos2.a[2942](config_982_4c3.o)'

On 10/16/2025 3:53 PM, Richard (Rikki) Andrew Cattermole wrote:
> Given that I just got an email stating that osx-13 runners is going away and to upgrade to osx-15-intel its probably a good idea to upgrade and see if that solves it.
> 
> We will be skipping osx-14 for the runner as it only comes in -large variant and that's paid I think.

Yes, concentrate on supporting the latest osx.