<testcase> # This test is quite timing dependent and tricky to set up. The time line of # test operations looks like this: # # 1. curl sends a PUT request with Expect: 100-continue and waits only 1 msec # for a 100 response. # 2. The HTTP server accepts the connection but waits 500 msec before acting # on the request. # 3. curl doesn't receive the expected 100 response before its timeout expires, # so it starts sending the body. It is throttled by a --limit-rate, so it # sends the first 64 KiB then stops for 1000 msec due to this # throttling. # 4. The server sends its 417 response while curl is throttled. # 5. curl responds to this 417 response by closing the connection (because it # has a half-completed response outstanding) and starting a new one. This # new request does not have an Expect: header so it is sent without delay. # It's still throttled, however, so it takes about 16 seconds to finish # sending. # 6. The server receives the response and this time acks it with 200. # # Because of the timing sensitivity (scheduling delays of 500 msec can cause # the test to fail), this test is marked flaky to avoid it being run in the CI # builds which are often run on overloaded servers. # Increasing the --limit-rate would decrease the test time, but at the cost of # becoming even more sensitive to delays (going from 500 msec to 250 msec or # less of accepted delay before failure). Adding a --speed-time would increase # the 1 second delay between writes to longer, but it would also increase the # total time needed by the test, which is already quite high. # # The assumption in step 3 is also broken on NetBSD 9.3, OpenBSD 7.3 and # Solaris 10 as they only usually send about half the requested amount of data # (see https://curl.se/mail/lib-2023-09/0021.html). <info> <keywords> HTTP HTTP PUT Expect flaky timing-dependent </keywords> </info> # Server-side <reply> # 417 means the server didn't like the Expect header <data> HTTP/1.1 417 BAD swsbounce Date: Tue, 09 Nov 2010 14:49:00 GMT Server: test-server/fake Content-Length: 0 </data> <data1> HTTP/1.1 200 OK Date: Tue, 09 Nov 2010 14:49:00 GMT Server: test-server/fake Content-Length: 10 blablabla </data1> <datacheck> HTTP/1.1 417 BAD swsbounce Date: Tue, 09 Nov 2010 14:49:00 GMT Server: test-server/fake Content-Length: 0 HTTP/1.1 200 OK Date: Tue, 09 Nov 2010 14:49:00 GMT Server: test-server/fake Content-Length: 10 blablabla </datacheck> <servercmd> no-expect delay: 500 connection-monitor </servercmd> </reply> # Client-side <client> <server> http </server> <name> HTTP PUT with Expect: 100-continue and 417 response during upload </name> <command> http://%HOSTIP:%HTTPPORT/we/want/%TESTNUMBER -T %LOGDIR/test%TESTNUMBER.txt --limit-rate 64K --expect100-timeout 0.001 </command> <precheck> perl -e "print 'Test does not work on this BSD system' if ( $^O eq 'netbsd' || $^O eq 'openbsd' || ($^O eq 'solaris' && qx/uname -r/ * 100 <= 510));" </precheck> # Must be large enough to trigger curl's automatic 100-continue behaviour <file name="%LOGDIR/test%TESTNUMBER.txt"> %repeat[132 x S]%%repeat[16462 x xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx%0a]% </file> </client> # Verify data after the test has been "shot" <verify> <protocol> PUT /we/want/%TESTNUMBER HTTP/1.1 Host: %HOSTIP:%HTTPPORT User-Agent: curl/%VERSION Accept: */* Content-Length: 1053701 Expect: 100-continue %repeat[132 x S]%%repeat[1021 x xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx%0a]%%repeat[60 x x]%[DISCONNECT] PUT /we/want/%TESTNUMBER HTTP/1.1 Host: %HOSTIP:%HTTPPORT User-Agent: curl/%VERSION Accept: */* Content-Length: 1053701 %repeat[132 x S]%%repeat[16462 x xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx%0a]% [DISCONNECT] </protocol> </verify> </testcase>