Bug 184503 – [GTK] imported/w3c/web-platform-tests/fetch/api/redirect/redirect-method.html fails

• Component: WebKitGTK
• Status: RESOLVED FIXED

[Alicia Boya García]

The following tests fail on GTK since W3C tests were updated in r230330:

imported/w3c/web-platform-tests/fetch/api/redirect/redirect-method.html
imported/w3c/web-platform-tests/fetch/api/redirect/redirect-method-worker.html


--- /home/slave/webkitgtk/gtk-linux-64-release-tests/build/layout-test-results/imported/w3c/web-platform-tests/fetch/api/redirect/redirect-method-worker-expected.txt
+++ /home/slave/webkitgtk/gtk-linux-64-release-tests/build/layout-test-results/imported/w3c/web-platform-tests/fetch/api/redirect/redirect-method-worker-actual.txt
@@ -10,6 +10,7 @@
 PASS Redirect 303 with POST 
 PASS Redirect 303 with HEAD 
 PASS Redirect 307 with GET 
-PASS Redirect 307 with POST 
+PASS Redirect 307 with POST (string body) 
+FAIL Redirect 307 with POST (blob body) assert_equals: request body expected "it's me the blob! and more blob!" but got ""
 PASS Redirect 307 with HEAD

[Rob Buis]

This passes for GTK. Close?

[Michael Catanzaro]

Not yet, there are still failure expectations in platform/gtk/TestExpectations.

[Rob Buis]

Created attachment 347530 [details]
Patch

[Michael Catanzaro]

Comment on attachment 347530 [details]
Patch

View in context: https://bugs.webkit.org/attachment.cgi?id=347530&action=review

All three tests are still failing on the GTK release bot:

--- /home/slave/webkitgtk/gtk-linux-64-release-tests/build/layout-test-results/imported/w3c/web-platform-tests/fetch/api/redirect/redirect-method-expected.txt
+++ /home/slave/webkitgtk/gtk-linux-64-release-tests/build/layout-test-results/imported/w3c/web-platform-tests/fetch/api/redirect/redirect-method-actual.txt
@@ -10,6 +10,7 @@
 PASS Redirect 303 with POST 
 PASS Redirect 303 with HEAD 
 PASS Redirect 307 with GET 
-PASS Redirect 307 with POST 
+PASS Redirect 307 with POST (string body) 
+FAIL Redirect 307 with POST (blob body) assert_equals: request body expected "it's me the blob! and more blob!" but got ""
 PASS Redirect 307 with HEAD 

--- /home/slave/webkitgtk/gtk-linux-64-release-tests/build/layout-test-results/imported/w3c/web-platform-tests/fetch/api/redirect/redirect-method-worker-expected.txt
+++ /home/slave/webkitgtk/gtk-linux-64-release-tests/build/layout-test-results/imported/w3c/web-platform-tests/fetch/api/redirect/redirect-method-worker-actual.txt
@@ -10,6 +10,7 @@
 PASS Redirect 303 with POST 
 PASS Redirect 303 with HEAD 
 PASS Redirect 307 with GET 
-PASS Redirect 307 with POST 
+PASS Redirect 307 with POST (string body) 
+FAIL Redirect 307 with POST (blob body) assert_equals: request body expected "it's me the blob! and more blob!" but got ""
 PASS Redirect 307 with HEAD 
 

--- /home/slave/webkitgtk/gtk-linux-64-release-tests/build/layout-test-results/imported/w3c/web-platform-tests/fetch/api/request/request-keepalive-quota-expected.txt
+++ /home/slave/webkitgtk/gtk-linux-64-release-tests/build/layout-test-results/imported/w3c/web-platform-tests/fetch/api/request/request-keepalive-quota-actual.txt
@@ -9,6 +9,6 @@
 PASS A Keep-Alive fetch() with a body at the Quota Limit should succeed. 
 PASS A Keep-Alive fetch() with a body over the Quota Limit should reject. 
 PASS A Keep-Alive fetch() should return its allocated Quota upon promise resolution. 
-PASS A Keep-Alive fetch() should return only its allocated Quota upon promise resolution. 
+FAIL A Keep-Alive fetch() should return only its allocated Quota upon promise resolution. assert_unreached: Should have rejected: undefined Reached unreachable code
 PASS A Keep-Alive fetch() should not be allowed if the Quota is used up. 
 

I think you've perhaps made the mistake of trying to run the tests locally. The results when run locally often do not match the test bots. Generally this occurs when some dependency is missing from the JHBuild environment. Anyway, our expectations need to match what the bots are reporting. If you've been following today's conversation in our internal Matrix, we've been discussing switching the bots over to run the tests inside a flatpak environment, which should for the first time allow for consistent test results when running locally. Hopefully this can happen in the next day or two....

> LayoutTests/ChangeLog:8
> +        Unskip these tests since they pass now.

Well there is a difference between [ Skip ], which actually prevents the tests from running, and [ Failure ], which runs the tests and verifies that the test does not match the expected result.

[Rob Buis]

(In reply to Michael Catanzaro from comment #4)
> I think you've perhaps made the mistake of trying to run the tests locally.

Right! I kind of missed the /home/slave/ hint. Ok I'll wait for the flatpack change.

[Rob Buis]

Created attachment 356180 [details]
Patch

[Rob Buis]

This was fixed in the gardening process.