Contents of segment at SHA256 digest offset 0xb0 are not all zero when building with protobuf for c++ (IDFGH-5417)

[MarcGeh]

Environment

• Development Kit: [ESP32-Wrover-Kit]
• Kit version: [v3]
• Module or chip used: [ESP32-WROVER]
• IDF version: v4.2.1
• Build System: [idf.py]
• Compiler version: xtensa-esp32-elf-gcc (crosstool-NG esp-2020r3) 8.4.0
• Operating System: [Linux]
• Using an IDE?: [No]
• Power Supply: [USB]

Problem Description

Hey guys, I've been trying to get protobuf c++ to run on the esp32. I created a protobuf component with all the source files from the protobuf git (version 3.15.2).

When running 'idf.py build' I get this error:

A fatal error occurred: Contents of segment at SHA256 digest offset 0xb0 are not all zero. Refusing to overwrite.

I have no idea what is causing this and why it only happens when I compile with the protobuf source files. When I comment out the lines 2056 - 2058 of esp-idf/components/esptool_py/esptool/esptool.py the binary is generated and I can run it on my dev-kit. However, there are some warnings during startup.

Since I really need the c++ implementation of protobuf to run on the esp32 and I have no idea how to fix this issue I thought I might get help here.

I created an example project to reproduce the issue.

Expected Behavior

Generate a binary image which can be flashed onto the target.

Actual Behavior

Generating binary image is aborted because of this error:

A fatal error occurred: Contents of segment at SHA256 digest offset 0xb0 are not all zero. Refusing to overwrite.

Steps to reproduce

Go and checkout the git repository I created for this: https://github.com/MarcGeh/esp32-cpp-protobuf
and build the project. The log output if the binary is generated with the empty check for the SHA256 digest commented out can be seen in the readme.

[o-marshmallow]

Hello @MarcGeh !
Thank you for your report, this is a bug that occurs when the alignment of the text or rodata section is changed because of the presence of an explicit alignment attribute within the code. In your repo, it happens because of the following piece of code:

PROTOBUF_FUNC_ALIGN(32)
void* Arena::AllocateAlignedNoHook(size_t n) { 
    return impl_.AllocateAligned(n, nullptr);
}

in arena.cc

Which is defined as:

// Function alignment attribute introduced in gcc 4.3
#define PROTOBUF_FUNC_ALIGN(bytes) __attribute__((aligned(bytes)))

in port_def.inc file

Fortunately, this bug has been fixed on main, release/4.3 and release/v4.2 branches. So you have multiple solutions:

• Switch to release/4.2 or newer branch
• Wait for a v4.2.2 release which will include this fix
• Patch locally your v4.2.1 release with the following commands:

# Go to ESP-IDF repo and make sure v4.2.1 is checked out
cd /path/to/esp-idf
git checkout v4.2.1

# Fetch the project to be sure to have the fix
git fetch

# Apply the newer commits that fix that bug
git cherry-pick 974b8dd4c40cc38c451089eeb9bf90c6dc22a22d af7b21851e9d5a7d1edc128c04f2e307bd2f9d98

# Go to esptool project to apply its fixes
cd components/esptool_py/esptool
git fetch
git cherry-pick 5619ecd723fd5c09d11d54f4768305ba45a341b0 627222a55d97b9886bc882672f07a5bf58d3f9b4

# You may experience a conflict here, no problem, the file in question is only used for testing, it can be ignored:
git reset test/test_imagegen.py
git checkout test/test_imagegen.py
git cherry-pick --continue

# Go back to your project to compile it 

[MarcGeh]

Hi @o-marshmallow
Thank you very much for the quick response and the clear and easy to understand solution! I can confirm that the binary is now generated after patching my esp-idf v4.2.1 locally like you have described. I will go ahead and close this issue.