Building eptri demo fails

[greatscottgadgets]

Building eptri demo fails

[ktemkin]

Sounds like something changed in one of the upstreams; probably. I'll check it out soon.

[kgugala]

@ktemkin did you have a chance to take a look at this?

[ktemkin]

I haven't come back around to looking at it; mostly because I've been laser focused on other things. I can take a look sometime this or next week, though. Poke me if I don't. :)

[ktemkin]

Just tested with the environment I already had installed without issue; it's using:

minerva => 0.1
nmigen-soc => 0.1.dev30+g425692a
lambdasoc =>  0.1.dev21+g95a97de
riscv64-unknown-elf-gcc => 9.2.0
riscv64-unknown-elf-newlib => 3.3.0

I'm guessing upgrading one of those python packages is going to break things -- trying that now. :)

[ktemkin]

Hmm; even after upgrading everything to its latest, I'm still not having issues:

minerva => 0.1 [but now installed from master]
nmigen-soc => 0.1.dev43+gecfad4d
lambdasoc =>  0.1.dev25+g7381b47
riscv64-unknown-elf-gcc => 9.2.0
riscv64-unknown-elf-newlib => 3.3.0

I'm particularly confused in that you're winding up with no stdint.h include, as simplesoc should always include that in the header it generates:

https://github.com/greatscottgadgets/luna/blob/master/luna/gateware/soc/simplesoc.py#L420

Some things that might be helpful in debugging:

1. What steps are you running to build? The simplest way is probably just to run make program in the eptri example folder.
2. Can you attach the resources.h and soc.ld that are generated by these build steps?

[kgugala]

looks like nmigen-stdio was missing in my setup. The build process was actually showing that in the log, but did not stop on missing dependency. This ended up with empty resources.h and in the end the reported error. Adding nmigen-stdio to requirements.txt solves the problem

[antoinevg]

eptri example code has moved to the Cynthion repo: https://github.com/greatscottgadgets/cynthion/