Martin Kinkelin

#4625 fixed this partially - an already set-up MSVC environment for ARM targets should work now, but setting it up manually would require more work, e.g., in function https://github.com/dlang/dmd/blob/748fab1e02b9f9e8980833b79e9448937bfecfab/compiler/src/dmd/vsoptions.d#L380.

Is this hack still required at Weka? The `needsCodegen()` stack pressure should have been improved significantly for a while now, in https://github.com/dlang/dmd/pull/14787.

This is a subset of https://github.com/ldc-developers/ldc/blob/e170ca51e673c258437915a8aff3feb31ea6802b/.cirrus.yml#L56-L64 (been a while since checking if there's been any improvements), which is more up-to-date than https://github.com/ldc-developers/ldc/issues/2153#issuecomment-626028446 from the AArch64 tracker issue. IIRC, the math...

> There's some inline assembly included by windows.h for 32-bit Windows that triggers this bug Note that similar problems wrt. missing (gcc) builtins and gcc-style inline assembly also happen on...

`-link-defaultlib-debug` links against unoptimized druntime/Phobos with debuginfos; I'd try that as first thing.

Thx for the report. Converted to D, compilable via `-preview=bitfields`: ```D struct BitField { uint _0 : 1; uint _1 : 1; uint _2 : 1; uint _3 : 1;...

Oh wait, the type-alloc size of a packed `` is apparently 5 bytes too! **Edit**: Well, it is 8 if unpacked.

Thx for the report. The problem isn't the icmp after the add (that's fine), but we need to load the lhs eagerly, not using it as lvalue and loading it...

Yeah as can be seen from the diff in https://github.com/ldc-developers/ldc/pull/4656, this wouldn't be doable for all cases anyway, most notably if an operand is a struct or class, since that...

Nice that you're hunting down the musl issues! :+1: - How's it looking, many failures still remaining or close to the finish line? Just wrt. v1.38.0 final that I'd like...