odb: fix createNewFlatNet flat/modnet name collision with BTerm

[openroad-ci]

Summary

• createNewFlatNet() now always names the new net after the BTerm (port name), and renames any colliding flat net or modnet to a unique avoidance name first.
• Fixes spurious assign statements and multi-driver errors in VerilogWriter output caused by remove_buffers leaving a BTerm on a mismatched net name with a stale modnet.
• Affected hierarchical designs like bp_fe_top where Kepler LEC rejected multi-driver outputs.

Root Cause

When remove_buffers merges two port-named nets (both having BTerms), the survivor selection defaults to the input net. This causes:

1. Flat net name mismatch: BTerm ends up on a net with a different port's name.
2. Stale modnet: remove_buffers merges flat nets but NOT modnets — the modnet retains the old port name.

createNewFlatNet() previously only handled case 1 (and only when names matched). VerilogWriter uses modnet names (via dbNetwork::net()) for submodule instance pin connections, so the stale modnet caused multi-driver even after flat net naming was corrected.

Fix

In createNewFlatNet():

• Always name the new net after the BTerm (regardless of current net name match).
• Rename the original flat net if it collides with the port name.
• Rename the modnet if it collides with the port name.
• Both renames share a single unique avoidance name generated via makeNewNetName().

Test plan

• All 37 InsertBuffer unit tests pass (including new Case33)
• New Case33 test: hierarchical design reproducing remove_buffers + insertBufferBeforeLoad multi-driver scenario
• Golden updates for Case1, Case3, Case9, Case10, Case12, Case13 (no more spurious assign)
• DEF golden updates for buffer_ports8, buffer_ports9, buffer_ports10 (BTerm net name now matches port name)

🤖 Generated with Claude Code

[github-actions]

clang-tidy review says "All clean, LGTM! 👍"

[gemini-code-assist]

Code Review

This pull request addresses a net naming collision issue that occurs during buffer insertion, particularly after remove_buffers has run. The fix ensures that when a new net is created before a BTerm, it is always named after the BTerm, and any existing flat nets or modnets with conflicting names are renamed to avoid collisions. The changes are well-supported by a new test case that reproduces the bug scenario. My review includes a suggestion to refactor a portion of the logic for improved readability and maintainability, which aligns with general best practices.

_{Note: Security Review did not run due to the size of the PR.}

[gemini-code-assist]

The logic for renaming the flat net and modnet to avoid collisions is correct, but it can be refactored for better readability and to make the intent clearer. You can determine if collisions exist for both the flat net and modnet first, and then perform the renames within a single conditional block. This avoids nested checks and separates collision detection from the renaming action.

    // Rename the original flat net and/or modnet if they have the same name as the port.
    const bool flat_net_collides = (bterm_name == block_->getBaseName(net_->getConstName()));
    dbModNet* load_mnet = bterm->getModNet();
    const bool mod_net_collides = (load_mnet
        && std::string_view{load_mnet->getConstName()} == bterm_name);

    if (flat_net_collides || mod_net_collides) {
      const std::string avoidance_name = make_avoidance_name();
      if (flat_net_collides) {
        net_->rename(avoidance_name.c_str());
      }
      if (mod_net_collides) {
        load_mnet->rename(avoidance_name.c_str());
      }
    }

[github-actions]

clang-tidy review says "All clean, LGTM! 👍"

[github-actions]

clang-tidy review says "All clean, LGTM! 👍"