Fix the issue with max_stanza_size #4197
Conversation
chrzaszcz
force-pushed
the
fix-max-stanza-size
branch
5 times, most recently
from
265937e
to
ba8a7a3
Compare
This comment was marked as outdated.
This comment was marked as outdated.
Codecov ReportAttention:
Additional details and impacted files@@ Coverage Diff @@
## master #4197 +/- ##
==========================================
+ Coverage 84.29% 84.33% +0.03%
==========================================
Files 549 549
Lines 33419 33419
==========================================
+ Hits 28171 28183 +12
+ Misses 5248 5236 -12 ☔ View full report in Codecov by Sentry. |
- Replace max_child_size with max_element_size - Replace 'child element too big' with 'element too big' - Remove unsupported and obsolete exml options: autoreset, start_tag
- Test with websockets, because they don't get chunked input and were failing before fixing exml. - Test the limit for stream opening tag as well.
- exml has the fix for max_element_size - escalus has the fix for handling the 'close' element in websockets
chrzaszcz
force-pushed
the
fix-max-stanza-size
branch
from
ba8a7a3
to
486f604
Compare
This comment was marked as outdated.
This comment was marked as outdated.
|
elasticsearch_and_cassandra_26 / elasticsearch_and_cassandra_mnesia / 309fb47 small_tests_25 / small_tests / 309fb47 small_tests_26 / small_tests / 309fb47 small_tests_26_arm64 / small_tests / 309fb47 ldap_mnesia_25 / ldap_mnesia / 309fb47 dynamic_domains_mysql_redis_26 / mysql_redis / 309fb47 ldap_mnesia_26 / ldap_mnesia / 309fb47 dynamic_domains_pgsql_mnesia_25 / pgsql_mnesia / 309fb47 sm_SUITE:parallel_manual_ack_freq_1:resume_session_state_send_message_without_ack{error,
{{assertion_failed,assert,is_presence,
{xmlel,<<"message">>,
[{<<"from">>,
<<"bob_resume_session_state_send_message_without_ack_2972@domain.example.com/escalus-default-resource">>},
{<<"to">>,
<<"alice_resume_session_state_send_message_without_ack_2979@domain.example.com">>},
{<<"type">>,<<"chat">>}],
[{xmlel,<<"body">>,[],[{xmlcdata,<<"msg-3">>}]}]},
"<message from='bob_resume_session_state_send_message_without_ack_2972@domain.example.com/escalus-default-resource' to='alice_resume_session_state_send_message_without_ack_2979@domain.example.com' type='chat'><body>msg-3</body></message>"},
[{escalus_new_assert,assert_true,2,
[{file,
"/home/circleci/project/big_tests/_build/default/lib/escalus/src/escalus_new_assert.erl"},
{line,84}]},
{sm_helper,initial_presence_step,2,
[{file,"/home/circleci/project/big_tests/tests/sm_helper.erl"},
{line,135}]},
{escalus_connection,connection_step,2,
[{file,
"/home/circleci/project/big_tests/_build/default/lib/escalus/src/escalus_connection.erl"},
{line,163}]},
{lists,foldl_1,3,[{file,"lists.erl"},{line,1355}]},
{escalus_connection,start,2,
[{file,
"/home/circleci/project/big_tests/_build/default/lib/escalus/src/escalus_connection.erl"},
{line,145}]},
{sm_helper,connect_spec,3,
[{file,"/home/circleci/project/big_tests/tests/sm_helper.erl"},
{line,153}]},
{sm_S...internal_mnesia_26 / internal_mnesia / 309fb47 dynamic_domains_pgsql_mnesia_26 / pgsql_mnesia / 309fb47 dynamic_domains_mssql_mnesia_26 / odbc_mssql_mnesia / 309fb47 pgsql_mnesia_25 / pgsql_mnesia / 309fb47 pgsql_cets_26 / pgsql_cets / 309fb47 pgsql_mnesia_26 / pgsql_mnesia / 309fb47 mysql_redis_26 / mysql_redis / 309fb47 mssql_mnesia_26 / odbc_mssql_mnesia / 309fb47 dynamic_domains_pgsql_mnesia_25 / pgsql_mnesia / 309fb47 |
The limit worked partially - only if the input was chunked, and not working properly for small limits. For websockets it wasn't working at all as reported in #4193.
The main fix is in the C++ code in
exml: esl/exml#68This PR also uses a fix for
escalus- I discovered an issue when adding a new test: esl/escalus#261These two PR's should be merged first. Then we could release them in hex, and update this PR.
Note: Latest escalus does not work with big tests anymore after merging esl/escalus#259, and this needs a fix before releasing the hex package.
Recommended to do in the future:
streamtag name verification in both regular C2S and websockets.