Martin Kinkelin
Martin Kinkelin
By completing https://github.com/ldc-developers/ldc/pull/3184, the transition to the ORCv2 API.
https://github.com/dlang/dmd/pull/13179 added that lowering to DMD's glue layer; it should probably be lowered in the frontend instead.
As of v2.098 (grep for `TODO: ImportC`): * initializer lists * ~~bit fields~~ (**edit**: added in https://github.com/ldc-developers/ldc/pull/4015) According dmd-testsuite tests have been disabled for now: * runnable/cstuff2.c * ~~runnable/bitfields*.c~~
E.g., it makes `-mtriple=mips64el-linux` incompatible with clang `-target mips64el-linux`, defaulting to 32-bit pointer size. If the option can be fully represented via `-mtriple` (e.g., apparently the case for MIPS), it...
Which requires calling different druntime hooks in the glue layer.
``` Testing forkgc2 /Users/runner/work/1/build/runtime/druntime-test-gc-debug/forkgc2 gnumake: *** [/Users/runner/work/1/build/runtime/druntime-test-gc-debug/forkgc2.done] Abort trap: 6 ``` Happened twice in a row yesterday with Azure image `20210626.1` after the tiny Travis change. Was green with `20210620.1`....
Based on https://issues.dlang.org/show_bug.cgi?id=14758#c15, DMD has improved wrt. unreferenced `TypeInfo`s ending up in the binary, while LDC still produces fat binaries (at least did in v1.0 according to the comment).
I.e., either port the codegen stuff from dlang/dmd#9282, or implement it differently in the frontend to require less codegen changes (Iain isn't happy about this at all: dlang/dmd#9702). Then revert...
`std.math` was split into a package with v2.097. This might imply that some little helpers (`isNaN` etc.) aren't inlined any longer without LTO (incl. Phobos), thus reducing performance of some...
As per title. In our case at work, standalone `dynamicLibrary` targets linking all dub deps statically now get dozens of static .lib files copied to their output directory. This didn't...