nimble build errors on Linux on jtv/v2 branch

[ee7]

In the GitHub Actions environment, when trying to run nimble test or nim r tests/test.nim:

/home/runner/work/con4m/con4m/tests/test.nim(3, 18) Error: undeclared identifier: 'toAnsiCode'
candidates (edit distance, scope distance); see '--spellSuggest': 
 (6, 4): 'AtomicNodes'
 (6, 4): 'HtmlNode'
 (6, 4): 'ItalicOff'
       Tip: 6 messages have been suppressed, use --verbose to show them.
tools.nim(36)            doCmd

    Error:  Execution failed with exit code 1
        ... Command: /home/runner/work/con4m/con4m/nimdir/bin/nim c --noNimblePath -d:NimblePkgVersion=0.2.0 --path:/home/runner/.nimble/pkgs2/nimutils-0.2.1-4689d581a064f06b3dbdcb2e60175a93c36797dd --path:/home/runner/.nimble/pkgs2/unicodedb-0.12.0-445[24](https://github.com/crashappsec/con4m/actions/runs/7958493760/job/21723477313#step:5:25)16471e2fe87[26](https://github.com/crashappsec/con4m/actions/runs/7958493760/job/21723477313#step:5:27)eb6070ed6ea7368171cc09 --hints:off -r --path:. /home/runner/work/con4m/con4m/tests/test

[viega]

Did you read the con4m doc I sent around yesterday? I wouldn't bother w the dev branch as nothing is the same (including the tests) and that should go away as quickly as we can manage the chalk integration for the v2 branch.

[ee7]

Did you read the con4m doc I sent around yesterday?

Yes. And I saw the bit about the new test harness.

I wouldn't bother w the dev branch

OK. Are things in a state where we can work towards getting tests running in CI for the v2 branch?

And is it expected that I currently get an error when running nimble build on the v2 branch?

$ git log -n1 --oneline
a59a292 (HEAD -> jtv/v2, crashappsec/jtv/v2) Remove the old c4m directory, it's not used.
$ nimble build
...
/tmp/nimble_23954/githubcom_crashappsecnimutils_#1b16dfecc54bda0d6492f96dfe4305bad68ec80d/nimutils/c/strcontainer.c: In function ‘c4str_from_file’:
/tmp/nimble_23954/githubcom_crashappsecnimutils_#1b16dfecc54bda0d6492f96dfe4305bad68ec80d/nimutils/c/strcontainer.c:65:34: error: ‘O_EXLOCK’ undeclared (first use in this function); did you mean ‘O_EXEC’?
   65 |     int fd = open(name, O_RDONLY|O_EXLOCK);
      |                                  ^~~~~~~~
      |                                  O_EXEC
/tmp/nimble_23954/githubcom_crashappsecnimutils_#1b16dfecc54bda0d6492f96dfe4305bad68ec80d/nimutils/c/strcontainer.c:65:34: note: each undeclared identifier is reported only once for each function it appears in

Edit: same error in CI for PR #121 (which I rebased on top of the v2 branch). O_EXLOCK and O_SHLOCK aren't portable, but I see this is recent code.

[ee7]

I updated #120 with crashappsec/nimutils@74130e3 and the latest v2 branch changes.

The next failure for nimble build on Linux in CI:

/usr/bin/ld: /home/runner/.cache/nim/test_r/@merr.nim.c.o: in function `print_con4m_trace':
@merr.nim.c:(.text+0xb9): undefined reference to `getStackTrace'
/usr/bin/ld: /home/runner/.cache/nim/test_r/@merr.nim.c.o: in function `customValidationError__err_u1570':
@merr.nim.c:(.text+0x31c1): undefined reference to `find_string_at'
/usr/bin/ld: /home/runner/.cache/nim/test_r/@[email protected]: in function `foreign_z_call':
@[email protected]:(.text+0x527): undefined reference to `run_0c00l_vm'
collect2: error: ld returned 1 exit status

[viega]

(In my last commit) I did what I could without having time to test on Linux to remove potential linker errors. The problem is, I see no reason why run_0c00l_vm wouldn't have linked, so might need someone else to address since I won't have access to my linux box.

[ee7]

(In my last commit) I did what I could without having time to test on Linux to remove potential linker errors. The problem is, I see no reason why run_0c00l_vm wouldn't have linked, so might need someone else to address since I won't have access to my linux box.

For visibility to others: it was because macOS builds enabled LTO, which caused files to be compiled that weren't compiled on Linux. Resolved in #128.