Today work has focused on a couple of things; Working through a new issue that’s popped up recently in nix-rust/nix, reviewing the implementation plan for the ctest extension patchset, and looking some more into Idris.
I started off by looking into a new ping I had gotten on the nix-rust/nix repo. This is a crate that provides safe bindings around rust-lang/libc symbols. They were experimenting with changes we are yet to release.
Their tests were failing on a certain set of signal.h-related wrappers, so I looked through the commit history in the libc-0.2 branch. I found there that my recent patch completing the siginfo_t definition introduced a regression.
In the PR that I opened, I replaced two untagged unions for regular records. This, of course, is fatal for siginfo_t’s sifields. Reading from it after crossing an FFI boundary gets you garbage, if not immediate UB.
Luckily, those changes, even though backported from the 1.0-tracking main branch, have not yet been released. I opened a new PR fixing this, and reported on this as well back in the nix-rust/nix thread.
They cherry-picked the commit from my branch and it seems the tests run just fine now.
The next thing I did was to submit the follow-up PR to the afore-mentioned siginfo_t patchset, which the reviewer requested to remove skips in the build script. Unfortunately, we could only remove the Linux skip.
Then I looked through the ctest extension patchset. I have not touched that since last week, though I recall that I had reached the point where I needed to implement item resolution at the module level from parsed use statements.
The plan back then was to perform one more pass after all items had been recursively parsed (including modules.) This pass would use the parsed FfiItems, and clone parsed items throughout modules by detecting reexports.
I still think this is the easiest way out, though once I started looking into it, I realized I was missing LSP assistance. After my recent workflow switch, the editor I am using is not LSP-enabled.
This prompted me to also spend some time today looking into setting that up. That is still not done, but I expect it to at least get me go-to-definition and other “basic” goodies by tomorrow.
Then I also looked into the Idris documentation, and worked through the first few sections of the crash course. I also finished up compiling pack, the toolchain and package manager for Idris.
I have yet to actually write some code in it, but I would like to first finish setting up LSP on my editor. Apparently, the idris2-lsp is pretty nice.
None.
I expect to resume work on the ctest extension stuff by tomorrow, once I get LSP working on my editor. The afore mentioned implementation plan continues to be the way I’m envisioning it.
In terms of side-project work, I believe I’ll be working through the rest of the crash course and the community tutorial until the end of the week. It may very well take longer, though. Then I will work through problems in the OCaml repo.