I fully agree that the ongoing development of fir-dev has split community efforts in a very damaging way and we must sort this out swiftly if the project is to have a future.
Pragmatically, if code drop + post-commit review, is the only way to bring the community together then it has my support on the condition that future development is conducted on main so we won’t have to resolve a situation like this again.
Given that any code drop is going to be in the many KLOC range, I would be as happy with a single big code drop as a handful of smaller ones. I assume this will be committed as sequential patches on top of main rather than as a merge commit(?)
Is there any update on this? As far as I can remember development on fir-dev has been reviewed at least by one person also present on Phabricator. Of course, a second review would not hurt. Looking forward to have this upstreamed as soon as possible.
This is to follow up from the conversation on the last group call.
The code in the flang-compiler/f18-llvm-project:fir-dev branch (aka: fir-dev) implements a functional pathway (with two distinct and operational bridges) from the flang front-end (parser and semantics checks) for the Fortran 77 standard. Using our testing tools (as a driver substitute – the actual driver is a work in progress itself), one can compile Fortran 77 programs to object files, link them with the flang runtime, and have correct, operational executables.
Changes to MLIR?
The fir-dev branch presently has a few small changes directly in MLIR (files in the mlir/ directory) that need to be upstreamed and completed.
Just as an update, the point 3 from Eric’s email was addressed by the following patch: https://reviews.llvm.org/D94737