gpu: Remove --cuda_verify_compatibility flag.

CUDA tests run in reasonable time with this flag enabled as well.
Remove this flag (hence enabling it be default) so that we catch
any test compatibility issues on pre-submits.

PiperOrigin-RevId: 796618710

Author: ayushr2

.buildkite/release.yaml

@@ -85,7 +85,7 @@ steps:
       <<: *retry_settings
     commands:
       - make sudo TARGETS=//tools/gpu:main ARGS="install --latest" || cat /var/log/nvidia-installer.log
-      - make cuda-tests ARGS="--cuda_verify_compatibility=true"
+      - make cuda-tests
     agents:
       queue: gpu
   - label: ":batfish: CUDA 12.8 tests"
@@ -97,7 +97,7 @@ steps:
       <<: *retry_settings
     commands:
       - make sudo TARGETS=//tools/gpu:main ARGS="install --latest" || cat /var/log/nvidia-installer.log
-      - make cuda-12-8-tests ARGS="--cuda_verify_compatibility=true"
+      - make cuda-12-8-tests
     agents:
       queue: gpu
   - <<: *common

test/gpu/cuda/cuda.go

@@ -63,10 +63,9 @@ const (
 
 // Flags.
 var (
-	verifyCompatibility = flag.Bool("cuda_verify_compatibility", os.Getenv("GVISOR_TEST_CUDA_VERIFY_COMPATIBILITY") == "true", "whether to verify that all tests are marked as compatible")
-	logSuccessfulTests  = flag.Bool("cuda_log_successful_tests", false, "log console output of successful tests")
-	debug               = flag.Bool("cuda_test_debug", false, "log more data as the test is running")
-	containersPerCPU    = flag.Float64("cuda_containers_per_cpu", defaultContainersPerCPU, "number of parallel execution containers to spawn per CPU (floating point values allowed)")
+	logSuccessfulTests = flag.Bool("cuda_log_successful_tests", false, "log console output of successful tests")
+	debug              = flag.Bool("cuda_test_debug", false, "log more data as the test is running")
+	containersPerCPU   = flag.Float64("cuda_containers_per_cpu", defaultContainersPerCPU, "number of parallel execution containers to spawn per CPU (floating point values allowed)")
 )
 
 // InitFlags parses the flags and sets the test.parallel flag to the desired
@@ -524,9 +523,6 @@ func runSampleTest(ctx context.Context, t *testing.T, testName string, te *TestE
 		compat = &FullyCompatible{}
 	}
 	willFailReason := compat.WillFail(ctx, te)
-	if willFailReason != "" && !*verifyCompatibility {
-		return fmt.Sprintf("this test is expected to fail (%s) --cuda_verify_compatibility=true to verify compatibility)", willFailReason), nil
-	}
 	if skipReason, isAlwaysSkipped := args.AlwaysSkippedTests[testName]; isAlwaysSkipped {
 		return fmt.Sprintf("this test is always skipped (%v)", skipReason), nil
 	}
@@ -676,8 +672,7 @@ type RunCudaTestArgs struct {
 	// parallel with them.
 	ExclusiveTests map[string]struct{}
 
-	// AlwaysSkippedTests don't run at all, ever, and are not verified when
-	// --cuda_verify_compatibility is set.
+	// AlwaysSkippedTests don't run at all.
 	// Each test is mapped to a reason why it should be skipped.
 	AlwaysSkippedTests map[string]string
 

test/gpu/cuda_12_8_test.go

@@ -35,9 +35,6 @@ var flakyTests = map[string]struct{}{}
 
 var exclusiveTests = map[string]struct{}{}
 
-// alwaysSkippedTests don't run at all, ever, and are not verified when
-// --cuda_verify_compatibility is set.
-// Each test is mapped to a reason why it should be skipped.
 // TODO(zkoopmans): Enable these tests once they pass.
 var alwaysSkippedTests = map[string]string{
 	"4_CUDA_Libraries/cudaNvSci":           "TODO - Debug",