Intenal Error of "Multiple root modules" happens if one of sub modules is not inlined when creating a protect lib.

[veripoolbot]

---

Author Name: Yutetsu TAKATSUKASA (@yTakatsukasa)
Original Redmine Issue: 1616 from https://www.veripool.org

Original Assignee: Todd Strader (@toddstrader)

---

First of all, thank you very much for maintaining and evolving the great software.

I am exploring --protect-lib to shorten the verilator runtime for a large design.
When I tried, I encountered Intenal Error of "Multiple root modules".
I can not share the actual design, so I made a tiny case to see the error.
The attached file is a small HDL to show the error.

I ran verilator as the following.

$ verilator --protect-lib tbp --top-module to_be_protected  --cc ./to_be_protected.sv

The error message is

%Error: Internal Error: to_be_protected.sv:10: ../V3ProtectLib.cpp:81: Multiple root modules
module to_be_protected  (input wire in, output wire out);
        ^~~~~~~~~~~~~~~
                         ... See the manual and https://verilator.org for more assistance.

As far as I tried, non-inlined module seems to cause the internal error because one of the following setting is necessary.

• Adding /verilator no_inline_module/ comment
• -O0 option

I tried the following versions and saw the error.
stable branch tag:verilator_4_022
master branch 3d6e8e9

[veripoolbot]

---

Original Redmine Comment
Author Name: Todd Strader (@toddstrader)
Original Date: 2019-11-29T23:21:09Z

---

Thanks for the example. I'll take a look.

[veripoolbot]

---

Original Redmine Comment
Author Name: Todd Strader (@toddstrader)
Original Date: 2019-11-30T16:37:07Z

---

That check was overly aggressive when inlining. Here's my proposed solution:
https://github.com/toddstrader/verilator-dev/tree/protect-inline

Wilson: I'm guessing this sort of fix doesn't require a Changes entry. But please let me know if it does. Also, please let me know if this needs any changes before pushing.

[veripoolbot]

---

Original Redmine Comment
Author Name: Wilson Snyder (@wsnyder)
Original Date: 2019-12-01T10:58:10Z

---

Todd, thanks for fixing. Feel free to push, you just need to init m_foundTop = false in the constructor.

[veripoolbot]

---

Original Redmine Comment
Author Name: Todd Strader (@toddstrader)
Original Date: 2019-12-01T13:02:22Z

---

m_foundTop is already in the constructor initializer list. Is it fine there or should I move it to the constructor's implementation?

[veripoolbot]

---

Original Redmine Comment
Author Name: Wilson Snyder (@wsnyder)
Original Date: 2019-12-01T16:01:44Z

---

Sorry, missed that, looks fine.

[veripoolbot]

---

Original Redmine Comment
Author Name: Todd Strader (@toddstrader)
Original Date: 2019-12-02T13:35:24Z

---

Fixed in git towards 4.024.