Add missing dependency of database tests on backend

[hramrach]

Missing make file dependency:

git clone https://github.com/go-gitea/gitea.git
make -C gitea test-sqlite

go test -c code.gitea.io/gitea/tests/integration -o integrations.sqlite.test -tags ' sqlite sqlite_unlock_notify'
sed -e 's|{{REPO_TEST_DIR}}||g'
-e 's|{{TEST_LOGGER}}|test,file|g'
-e 's|{{TEST_TYPE}}|integration|g'
tests/sqlite.ini.tmpl > tests/sqlite.ini
GITEA_ROOT="/home/hramrach/gitea" GITEA_CONF=tests/sqlite.ini ./integrations.sqlite.test
Could not find gitea binary at /home/hramrach/gitea/gitea

make: *** [Makefile:506: test-sqlite] Error 1

[silverwind]

CI failures are unrelated and will be fixed with #36074.

[silverwind]

CI failures are unrelated and will be fixed with #36074.

Actually no, not entirely unrelated. go: -race requires cgo; enable cgo by setting CGO_ENABLED=1 is related.

[wxiaoguang]

Well, I don't think this "fix" is right and can be approved as the current implementation.

@silverwind @delvh , see the CI failures. There are more details.

As per @wxiaoguang

[hramrach]

Interesting.

Clearly the dependency is real because the workflow in question first builds backend, and then runs the test.

However, the test is run with environment variables set that fail the backend build.

My suggestion would be to drop the CGO_ENABLED=0 which would then make the backend buildable as part of the test but I am not expert on go and don't know what this was supposed to accomplish.

[hramrach]

I see. cgo, at least the version you use is broken on Arm due to bogus inline assembly.

[silverwind]

I think we may just need to add CGO_ENABLED = 1 after below line because the race detector depends on it and RACE_ENABLED should only ever be set during test runs:

gitea/Makefile

Line 93 in 46d7ade

GOTESTFLAGS += -race

[silverwind]

Timeout after 50m seems odd as the db tests usually take only 30m, I've restarted the failing job, maybe it was a flake.

[silverwind]

If the postgres test times out again, try raising this temporarily:

.github/workflows/pull-db-tests.yml:56:        timeout-minutes: 50

[wxiaoguang]

If the postgres test times out again, try raising this temporarily:

.github/workflows/pull-db-tests.yml:56:        timeout-minutes: 50

No, it shouldn't and don't do that. There must be a bug (eg: deadlock) if it timeouts.

[silverwind]

It's still valuable if we see the output of a full test run that is not cancelled, otherwise you have no full picture of the errors to work with. Will of course be reverted before merge.

[wxiaoguang]

It's still valuable if we see the output of a full test run that is not cancelled, otherwise you have no full picture of the errors to work with. Will of course be reverted before merge.

For testing-only purpose, you can try. But I guess it won't really help, because usually a test finishes in 20-30 minutes. So you already have the doubled time.

[silverwind]

Yeah i don't know why it's taking so long now, I assume it's related to RACE_ENABLED which makes go test a lot more slow. But to my knowledge, this flag should have already been active for CI tests.