Use -conf= to overwrite the default dmd config

[wilzbach]

-conf= should almost always be used in any of our Makefiles to avoid any issues with user-defined configuration files.

Note: there's currently an permananet failure on some CIs. See also #7838 (comment)

[dlang-bot]

Thanks for your pull request, @wilzbach!

Bugzilla references

Your PR doesn't reference any Bugzilla issue.

If your PR contains non-trivial changes, please reference a Bugzilla issue or create a manual changelog.

Testing this PR locally

If you don't have a local development environment setup, you can use Digger to test this PR:

dub fetch digger
dub run digger -- build "master + dmd#7846"

[JinShil]

-conf= is already set in the execution of each test. See

dmd/test/d_do_test.d

Line 677 in c55c9be

string command = format("%s -conf= -m%s -I%s %s %s -od%s -c %s %s", envData.dmd, envData.model, input_dir,

[marler8997]

-conf= is already set in the execution of each test...

I believe that only affects .d tests, not .sh tests. Setting it in the Makefile should affect ALL tests.

[JinShil]

Then, shouldn't this PR also remove conf= from d_do_test.d?

[marler8997]

Then, shouldn't this PR also remove conf= from d_do_test.d?

I can't think of a reason that d_do_test would need to redundantly set -conf=, so I suppose that should work.

[marler8997]

One thought (somewhat unrelated). Sometimes it's nice to be able to copy the command line and run it on it's own (add/remove arguments like -v) and the more arguments that depend on the Makefile setting up the environment, the more arguments you have to manually add back into the command line. Maybe redundancy is ok in this case.

[wilzbach]

It should be "${DFLAGS[@]}" if expansion to multiple arguments is wanted, but DFLAGS is already an environment variable, so this isn't needed.
BTW it might be worthwhile to set -m"${MODEL}" in dflags too to avoid accidentally forgetting -m"${MODEL}" when it matters in the bash scripts.

[wilzbach]

I can't think of a reason that d_do_test would need to redundantly set -conf=, so I suppose that should work.

I left it in d_do_test because the environment flags are printed on a failure.
So when a test failed, you get this nice output with all parameters that were used to run the file:

Test compilable/test10056.d failed.  The logged output:
../generated/linux/release/64/dmd -conf= -m64 -Icompilable  -fPIC  -odtest_results/compilable -oftest_results/compilable/test10056_0.o  -c compilable/test10056.d 

This isn't so important for the bash scripts, because you can't run them directly anyways and you see all variables used thanks to xtrace.

[marler8997]

I left it in d_do_test because the environment flags are printed on a failure.
So when a test failed, you get this nice output with all parameters that were used to run the file

It doesn't print the "environment variables" though...does it?

[wilzbach]

Nope and we shouldn't be depending on the user to set environment variables to repeat the run of a test. I know in theory we could do DFLAGS="..." generated..., but that I think the current behavior of d_do_test is rather nice.
BTW d_do_test enforces DFLAGS to be an empty string to avoid exactly these problems:

dmd/test/d_do_test.d

Line 603 in e1f2be6

throw new Exception("The DFLAGS test argument must be empty: It is '" ~ testArgs.dflags ~ "'");

So the "redundant" -conf= is really required.

[JinShil]

Then, shouldn't this PR also remove conf= from d_do_test.d?

Please address this.

[Geod24]

Ping @wilzbach

[RazvanN7]

The Makefile for running tests is deprecated. Closing this.