macOS compilation error: <endian.h> file not found

[npes-95]

This file isn't found when compiling under macOS. To compile under macOS, you need to modify the include in helpers/rlp_encoder.c to read:

#include <machine/endian.h>

Would this still be valid under Linux?

[lorepieri8]

Can you specify which build are you playing with and report the Error log?

[npes-95]

This is for the x86 build. The log reads:

../helpers/rlp_encoder.c:9:10: fatal error: 'endian.h' file not found
#include <endian.h>

[lorepieri8]

Did you installed solc and run the

pip install --user -r requirements.txt

(or pip3)

?

[npes-95]

Yep. I'll go over it again now to make sure everything is okay.

[npes-95]

Everything is okay as far as I can tell. Is the ABI compiler called using python or python3? My default python is 2.7.

[npes-95]

Apparently, you can include <sys/types.h> which will return the correct endian file for Linux/macOS.

https://stackoverflow.com/questions/27073837/fatal-error-endian-h-file-not-found

[npes-95]

Okay so I've done a bit of digging, the <machine/endian.h> file on macOS doesn't define the htobe64() method, so there would have to be some kind of shim to make the code work on macOS.

https://gist.github.com/yinyin/2027912

[npes-95]

I've just added the shim code to usr/local/include and the project compiles!

[pcppcp]

Hi @npes-95, we don't have many MacOS users so I am sorry that the build so wonky :) Great you made it work!

Regarding the endianess the shim code looks like a good solution - could you make a pull request for that?

Thanks!

[npes-95]

That's okay, I was considering giving up and setting up a dual boot for a minute!

So the shim code needs to go in the user's local/include, should I modify an existing build script to copy that in there? I reckon we should keep a copy of the shim code in the repo in case the gist is deleted, where would be the best place to put it?

[pcppcp]

I'd create a separate directory for the shim (i.e. helpers/macos) and then we can conditionally include that path for the MacOS builds.

helpers/CMakeLists.txt

if (${CMAKE_SYSTEM_NAME} MATCHES "Darwin")
    include_directories(helpers/macos)
endif()

[devtodev]

Hi! when I compiled for my desktop (Linux Fedora) all runs well, but when I change the target to a board like DBOARD=frdm_kl25z or DBOARD=nrf52840_pca10056 the endian.h file isn't found when run ninja. I tried to add <sys/types.h> but is the same.

[pcppcp]

@devtodev for this error, can you create a gist with an output of ninja -v

[devtodev]

Here it is https://gist.github.com/devtodev/ab7a30744a71bd83191d03ade64766ec

[pcppcp]

Thanks! Now I see where the problem is - I haven't noticed there is no endian.h for Zephyr. I'll create a shim file similar to one in #63 and then it should work.

[pcppcp]

that reminds me of the fact we desperately need CI builds for Zephyr target (#52) .. volunteers? :D

[pcppcp]

could you guys check out current master? It should compile now. (for zephyr builds)

[devtodev]

still is broken the zephyr compilation, here is the output https://gist.github.com/devtodev/b2edd806469eea457bbd33595fb16b60 I'll be working on it to fix it, if you have any advice please show me the light ;)

[devtodev]

Hi @pcppcp! I can compile for zephyr now, but to do that I made two "fixs". The first was fix wallet.c by comment the include "eth/account.h" and declare account_t struct, like this:

typedef struct {
    uint8_t *privkey;
    unsigned long nonce;
} account_t;

The correct solution will be add to the repo eth/account.h, I searched that file with:

git rev-list -n 1 HEAD -- eth/account.h

And the file never existed in this repo. Maybe when wallet.c was added, the developer forgot to add account.h to the repo.

The second fix was remove SHELL_REGISTER function call and their struct, because the struct not was correct filled. Do you know the correct way to fil this structure?

Well.. with this two fixes, the zephyr build compile fine, but to get the correct solution of this issues we need add eth/account.h and fill fine the struct of SHELL_REGISTER. Right?

[MohitKAgnihotri]

@devtodev I think I have the solution for the second question. We can register shell command as shown below.

SHELL_CREATE_STATIC_SUBCMD_SET(sub_crypto)
{
       /* Alphabetically sorted to ensure correct Tab autocompletion. */
       SHELL_CMD(pk2addr, NULL, "pk2addr", cmd_pk2addr),
       SHELL_SUBCMD_SET_END /* Array terminated. */
};

SHELL_CMD_REGISTER(crypto, &sub_crypto, "crypto eth", NULL);

[pcppcp]

@devtodev oops the account_t got lost somewhere. I've fixed it in PR #68

@MohitKumarAgniotri looks like you are using different version of Zephyr, I can't find any of the SHELL_* macros in the code. I'll update and take a look at that.

[pcppcp]

Yea I've been using Zephyr version from August so unsurprisingly it didn't compile for you :) Fixes in #69

[npes-95]

Hey guys I'm gonna close this issue, as the solution for this has been merged. If there's still other problems, make a separate issue for them.