Hi, I’ve encountered issues building Chromium while experimenting with the suggested (non-)thinLTO configurations. Here are the main challenges:
- Linker Errors: When attempting to build Chromium using various (non-)thinLTO setups, I ran into numerous linker errors that halted progress. A few examples:
ld.lld: error: (...): loop not unrolled: the optimizer was unable to perform the requested transformation; the transformation might be disabled or specified as part of an unsupported transformation ordering
ld.lld: error: section .llvmbc at 0x0 of size (...) exceeds available address space
ld.lld: error: File name too long
- Unicode Decode Error: Using a different configuration, I successfully built the project but encountered this error when extracting corpus:
UnicodeDecodeError: 'utf-8' codec can't decode byte 0xc3 in position (...): invalid continuation byte`
After changing the extract_ir.py script to force encoding to be latin-1, the first command line runs but in the end no files get converted.
Given that collecting the Default Trace relies on a complete build (my assumption for some missing libraries errors), resolving this is essential for continuing the demo.
Questions:
- Has there been a recently tested MLGO setup with Chromium that bypasses these issues?
- Would greatly appreciate any advice on resolving the linker errors and the UnicodeDecodeError.
Thanks in advance for your help!
Hi, I’ve encountered issues building Chromium while experimenting with the suggested (non-)thinLTO configurations. Here are the main challenges:
After changing the
extract_ir.pyscript to force encoding to be latin-1, the first command line runs but in the end no files get converted.Given that collecting the Default Trace relies on a complete build (my assumption for some missing libraries errors), resolving this is essential for continuing the demo.
Questions:
Thanks in advance for your help!