OBS service test failed since appending object API return ETag for object this time

[Ziy1-Tan]

Hey! I am working on #5693 for obs service. But I encountered some unexpected results and am not sure which solution is better.

The obs append API will return ETag for append object this time instead of entire object itself: https://support.huaweicloud.com/api-obs/obs_04_0088.html#section9.
It make test test_blocking_write_with_append_returns_metadata fail:

thread '<unnamed>' panicked at tests/behavior/async_write.rs:675:9:
assertion `left == right` failed
  left: Some("\"5ade25cb48b852add9aadb6f0e427ac5-2\"")
 right: Some("\"753ba0d90cf2306e3bb6e3543c1e7d98\"")
note: run with `RUST_BACKTRACE=1` environment variable to display a backtrace
test behavior::test_write_with_append_returns_metadata ... FAILED

failures:

---- behavior::test_write_with_append_returns_metadata ----
test panicked: assertion `left == right` failed
  left: Some("\"5ade25cb48b852add9aadb6f0e427ac5-2\"")
 right: Some("\"753ba0d90cf2306e3bb6e3543c1e7d98\"")


failures:
    behavior::test_write_with_append_returns_metadata

test result: FAILED. 0 passed; 1 failed; 0 ignored; 0 measured; 138 filtered out; finished in 0.95s

How do we cover this case, make it a TODO or add a special use case to cover it?

[Ziy1-Tan]

cc @Xuanwo @meteorgan .

[Xuanwo]

Yes. I think maybe it's time for us to remove all write_has_xxx capabilities.

[Xuanwo]

Tracked at #5860

[Ziy1-Tan]

Is there any relationship between removing write_has_xxx and covering obs test cases? It seems that the meta returned by obs append and stat is a special case, and the test cases cannot distinguish obs from other standard services.

[meteorgan]

Once we drop write_has_xxx, we can skip the etag returned by append from OBS and still obtain and test the etag returned by write.

[meteorgan]

Another similar issue: In OSS, PutObject returns content-md5, but AppendObject doesn't.