Flush gzip buffer when cancelling http response buffer

[tavplubix]

Changelog category (leave one):

• Bug Fix (user-visible misbehavior in an official stable release)

Changelog entry (a user-readable short description of the changes that goes into CHANGELOG.md):

Flush buffers when sending an error in the middle of compressed stream in HTTP interface.

Details

See #75175 (comment)

The problem is that ZlibDeflatingWriteBuffer::nextImpl() doesn't necessarily flush the buffer, we need to finalize it.

[clickhouse-gh]

Workflow [PR], commit [fcac704]

Summary: ❌

job_name	test_name	status	info	comment
Integration tests (amd_asan, old analyzer, 4/6)		failure
	test_scheduler_cpu_preemptive/test.py::test_downscaling[cpu-slot-preemption-timeout-1ms]	FAIL	cidb, flaky
	test_merges_memory_limit/test.py::test_memory_limit_success	FAIL	cidb
	test_backup_restore_on_cluster/test_huge_concurrent_restore.py::test_backup_restore_huge_cluster	FAIL	cidb
	test_replicated_merge_tree_compatibility/test.py::test_replicated_merge_tree_defaults_compatibility	FAIL	cidb
	OOM in dmesg	FAIL	cidb
Integration tests (amd_tsan, 3/6)		failure
	test_replication_without_zookeeper/test.py::test_startup_without_zookeeper	FAIL	cidb
Upgrade check (amd_debug)		failure
	Killed by signal (in clickhouse-server.log)	FAIL	cidb
	Fatal message in clickhouse-server.log (see fatal_messages.txt)	FAIL	cidb
	Killed by signal (output files)	FAIL	cidb
	Found signal in gdb.log	FAIL	cidb

[kavirajk]

@tavplubix FYI: Currently it also broken on other compressions like brotli

for example

echo "select throwIf(number > 10000000) + number check from numbers(10000005) format CSV" | curl "http://127.0.0.1:8123/?query=&enable_http_compression=1" -H "Accept-Encoding: br" --compressed -v --data-binary @- | tail -c 1024

is the intention is to fix all the broken compression in this PR or just gzip?

(it would be nice to write tests for all the compression supported by CH server?)

[tavplubix]

is the intention is to fix all the broken compression in this PR or just gzip?

all, it fixes brotli as well, I added it to the test

(it would be nice to write tests for all the compression supported by CH server?)

curl supports only deflate, gzip, br and zstd, so adding other compression methods to the test is not trivial, but I don't think it's necessary since the fix is generic

[kavirajk]

Thanks @tavplubix. make sense 👍

[kavirajk]

@tavplubix can we merge this? failing tests looks unrelated?