[SAOC 2025] Improve importC Weekly Update #2

Sep 29

Emmanuel

Sep 29

jmh530

Sep 29

Emmanuel

Sep 29

Richard (Rikki) Andrew Cattermole

Sep 29

Emmanuel

September 29

[SAOC 2025] Improve importC Weekly Update #2

Posted by Emmanuel

Permalink

Emmanuel

Permalink

Hi everyone,

This week, I had a long run with issue 20092 working to make compound literals fully work in importC and it has been successfully fixed and merged.

issue: https://github.com/dlang/dmd/issues/20092

The whole idea as discussed with my mentor was to create a temporary variable and then export it with the addressof operator during semantic analysis.

I did a lot of research into the expression semantics to fully understand how the compound literals would fit together. for local pointers, creating a temporary and making sure the variable declaration and the variable expression were exported with the addressof was enough.

for globals, it needed a little extra tinkering by making sure I push the symbol into the symbol table for the address to be picked up by the pointer.

PR : https://github.com/dlang/dmd/pull/21908#event-19961166923

int *p = &(int){0}; in importC should get you a valid pointer which you can work with and not necessarily creating an object before pointing it as defined by the C standard.

On 30/09/2025 2:23 AM, Emmanuel wrote: > On Monday, 29 September 2025 at 12:51:59 UTC, jmh530 wrote: >> On Monday, 29 September 2025 at 08:51:00 UTC, Emmanuel wrote: >>> Hi everyone, >>> >>> This week, I had a long run with issue 20092 working to make compound literals fully work in importC and it has been successfully fixed and merged. >>> >>> [...] >> >> Thanks. Do you know if the test suite includes C projects? > > Yes, I think. there are C files being run by DMD in compilable and runnable tests. > even in fail compilations as well. > > I don't really get what you mean by `C projects` but C files are being tested. Projects like sljit. We do this with D projects using the buildkite CI test. But not with C. This is one of the libraries I consider a requirement for ImportC to be minimally operational. https://github.com/dlang/dmd/issues/21891 https://github.com/dlang/dmd/issues/21890 ImportC is getting really close, but yes it would be nice if we could track this with actual C projects being tested against.

On Monday, 29 September 2025 at 13:42:27 UTC, Richard (Rikki) Andrew Cattermole wrote: > On 30/09/2025 2:23 AM, Emmanuel wrote: >> On Monday, 29 September 2025 at 12:51:59 UTC, jmh530 wrote: >>> [...] >> >> Yes, I think. there are C files being run by DMD in compilable and runnable tests. >> even in fail compilations as well. >> >> I don't really get what you mean by `C projects` but C files are being tested. > > Projects like sljit. We do this with D projects using the buildkite CI test. But not with C. > > This is one of the libraries I consider a requirement for ImportC to be minimally operational. > > https://github.com/dlang/dmd/issues/21891 > > https://github.com/dlang/dmd/issues/21890 > > ImportC is getting really close, but yes it would be nice if we could track this with actual C projects being tested against. Alright, I agree. I should definitely look into this.

Forums