This week marked an exciting checkpoint in my LFortran project — I began the actual compilation of Fortran Package Manager (FPM) using the LFortran compiler. After successful compilation of its dependencies like M_CLI2, jonquil, and others with some workarounds, I started tackling the core of FPM itself.

Process is same for this also, I started with creating workarounds. This effort led to the discovery of multiple frontend AST->ASR issues, which I translated into clear, minimal reproducing examples (MREs) which I will fix later. Overall this week was dedicated towards finding core issues in FPM modules that we need to tackle in order to compile it with LFortran.

Work done this week:

• Started compiling FPM core files:
  • Modules such as fpm_model, fpm_filesystem, fpm_package, and fpm_source_file etc. began compiling with some minor adjustments.
  • Several issues found regarding operator overloading, class(*) variable handling, external enum and inter dependent derived types.
  • Introduced workarounds, reated 8-10 MREs, fixed 2-3 of them.
  • Also I fixed 2-3 MREs which were created during MCLI2 compilation.
• Important MREs created this week:
  • Issue 7868: Handle struct used in another struct which is declared before in module
  • Issue 7869: Support mold argument of allocate for class type variables
  • Issue 7878: Private function of a module is exposed via use
  • Issue 7893: Handle overloaded function when one is elemental with element type and other has array type argument
• Pull Requests Submitted:
  • PR #7903: Compare derived type correctly while checking operator overloading
  • PR #7849: Get current type in visit_cast for select type stmts
  • PR #7867: Get external struct symbol correctly

Around 18 hours were dedicated to FPM compilation this week. While some modules are still blocked, the issues are clear and tracked. Multiple MREs were created with specific focus on runtime polymorphism, operator overloading, and interface resolution in type-bound procedures. The collaborative effort between myself and the LFortran team has made debugging quicker and more targeted.

A big thanks again to Ondřej Čertík and Pranav Goswami for patiently reviewing issues and pull requests. Thanks to LFortran team to helping me wherever needed.

Plan for next week:

• Resolve major blocking MREs for FPM compilation.
• Create new MREs once unblocked from previous ones.
• Try to compile FPM untill end of next week. So that we have enough time to test all codes.