[Flatpak SDK] Switch runtime to 21.08 SDK

• Component: WebKitGTK
• Status: RESOLVED FIXED

[Philippe Normand]

- The UserFlatpak update is automatic, no specific action needed
- Bots will still need clean builds though (remove WebKitBuild/{GTK,WPE})

[Philippe Normand]

> - Bots will still need clean builds though (remove WebKitBuild/{GTK,WPE})

Actually, we can trigger cmake to reconfigure, by removing CMakeCache.txt files in build dirs during the SDK upgrade, so no action would be needed on bots, hopefully.

[Philippe Normand]

+ Yusuke San in CC

I'm seeing new test262 test results when using ICU 69.1 here:

Settings:
Test262 Dir: ../../../../app/webkit/JSTests/test262
JSC: ../../../../app/webkit/WebKitBuild/Release/bin/jsc
Child Processes: 24
DYLD_FRAMEWORK_PATH: /app/webkit/WebKitBuild/Release/bin
Config file: ../../../../app/webkit/JSTests/test262/config.yaml
Expectations file: ../../../../app/webkit/JSTests/test262/expectations.yaml
---

NEW PASS test/intl402/DateTimeFormat/prototype/formatRange/fractionalSecondDigits.js (strict mode)
NEW PASS test/intl402/DateTimeFormat/prototype/formatRange/en-US.js (strict mode)
NEW PASS test/intl402/DateTimeFormat/prototype/formatRange/en-US.js (default)
NEW PASS test/intl402/DateTimeFormat/prototype/formatRange/fractionalSecondDigits.js (default)
! NEW FAIL test/intl402/Intl/getCanonicalLocales/preferred-grandfathered.js (strict mode)
Exception: Test262Error: Expected SameValue(«cel-gaulish», «xtg») to be true

! NEW FAIL test/intl402/Intl/getCanonicalLocales/preferred-grandfathered.js (default)
Exception: Test262Error: Expected SameValue(«cel-gaulish», «xtg») to be true

NEW PASS test/intl402/Intl/getCanonicalLocales/non-iana-canon.js (strict mode)
! NEW FAIL test/intl402/Locale/getters-grandfathered.js (strict mode)
Exception: Test262Error: Expected SameValue(«cel-gaulish», «xtg») to be true

! NEW FAIL test/intl402/Locale/getters-grandfathered.js (default)
Exception: Test262Error: Expected SameValue(«cel-gaulish», «xtg») to be true

NEW PASS test/intl402/Intl/getCanonicalLocales/non-iana-canon.js (default)
! NEW FAIL test/intl402/Locale/extensions-grandfathered.js (strict mode)
Exception: Test262Error: Expected SameValue(«fr-Cyrl-FR-gaulish-u-nu-latn», «fr-Cyrl-FR-u-nu-latn») to be true

! NEW FAIL test/intl402/Locale/likely-subtags-grandfathered.js (strict mode)
Exception: Test262Error: Expected SameValue(«cel-gaulish», «xtg») to be true

! NEW FAIL test/intl402/Locale/extensions-grandfathered.js (default)
Exception: Test262Error: Expected SameValue(«fr-Cyrl-FR-gaulish-u-nu-latn», «fr-Cyrl-FR-u-nu-latn») to be true

! NEW FAIL test/intl402/Locale/likely-subtags-grandfathered.js (default)
Exception: Test262Error: Expected SameValue(«cel-gaulish», «xtg») to be true

NEW PASS test/intl402/Locale/prototype/minimize/removing-likely-subtags-first-adds-likely-subtags.js (strict mode)
NEW PASS test/intl402/Locale/prototype/minimize/removing-likely-subtags-first-adds-likely-subtags.js (default)

78016 tests run
1004 test files skipped
1120 tests failed in total
8 tests newly fail
8 tests newly pass
Saved all the results in /var/home/phil/WebKit/test262-results/results.yaml
Summarizing results...
See the summaries and results in the /var/home/phil/WebKit/test262-results.


I wonder how to address this, I can't find port-specific expectations for test262. Can you advise on what should be done please?

[Philippe Normand]

Same question for new JSC intl-locale failures ;)

[Lauro Moura]

While testing, I've compiled a table of the packages upgraded (through pkg-config versions): https://gist.github.com/lauromoura/7f97b15cb7c06189956d46242b918194

Still getting some errors, though (didn't dive too deep yet, looks like some non-unified errors):

In file included from /app/webkit/Source/WebKit/NetworkProcess/webrtc/NetworkRTCMonitor.h:30,
                 from /app/webkit/WebKitBuild/Release/DerivedSources/WebKit/NetworkRTCMonitorMessageReceiver.cpp:27:
/app/webkit/Source/WebKit/Shared/RTCNetwork.h:49:17: error: ‘optional’ in namespace ‘std’ does not name a template type
   49 |     static std::optional<RTCNetwork> decode(IPC::Decoder&);
      |                 ^~~~~~~~
/app/webkit/Source/WebKit/Shared/RTCNetwork.h:34:1: note: ‘std::optional’ is defined in header ‘<optional>’; did you forget to ‘#include <optional>’?
   33 | #include <wtf/Forward.h>
  +++ |+#include <optional>
   34 | 
/app/webkit/Source/WebKit/Shared/RTCNetwork.h:56:21: error: ‘optional’ in namespace ‘std’ does not name a template type
   56 |         static std::optional<IPAddress> decode(IPC::Decoder&);
      |                     ^~~~~~~~
/app/webkit/Source/WebKit/Shared/RTCNetwork.h:56:16: note: ‘std::optional’ is defined in header ‘<optional>’; did you forget to ‘#include <optional>’?
   56 |         static std::optional<IPAddress> decode(IPC::Decoder&);
      |                ^~~
/app/webkit/Source/WebKit/Shared/RTCNetwork.h:68:21: error: ‘optional’ in namespace ‘std’ does not name a template type
   68 |         static std::optional<SocketAddress> decode(IPC::Decoder&);
      |                     ^~~~~~~~
/app/webkit/Source/WebKit/Shared/RTCNetwork.h:68:16: note: ‘std::optional’ is defined in header ‘<optional>’; did you forget to ‘#include <optional>’?
   68 |         static std::optional<SocketAddress> decode(IPC::Decoder&);
      |                ^~~
[1459/2037] Building CXX object Source/WebKit/CMakeFiles/WebKit.dir/__/__/DerivedSources/WebKit/WebPageProxyMessageReceiver.cpp.o

[Philippe Normand]

Are you using my branch? The wip commit fixes this optional error.

[Philippe Normand]

Created attachment 438658 [details]
Patch

[Michael Catanzaro]

Comment on attachment 438658 [details]
Patch

These test expectations changes are a little weird. On the one hand, there are fewer than I would have expected. On the other hand, the ones that are there are *weird*. I wonder what changed, but I also know it would be pretty hard to find out.

[Philippe Normand]

Which ones are weird?

It's not the first time a libxml2 update changes test expectations, fwiw.

[Philippe Normand]

There is an issue with the GTK API tests, possibly related with xvfb... I'll check.

[Philippe Normand]

(In reply to Philippe Normand from comment #9)
> There is an issue with the GTK API tests, possibly related with xvfb... I'll
> check.

Yes... Carlos, there is a weird issue again with xvfb / subprocess.

If I redirect the stderr process to /dev/null the test runner is running OK. Otherwise it runs a few tests and hangs forever. My WIP diff is http://sprunge.us/3gJMX6

The SDK runtime has been updated from python 3.8 to python 3.9. Could this be related?

[Philippe Normand]

While the test runner is hanged up I see a weird process from the host with `ps aux`:

/usr/bin/xkbcomp -w 1 -R/usr/share/X11/xkb -xkm - -em1 The XKEYBOARD keymap compiler (xkbcomp) reports: -emp >  -eml Errors from xkbcomp are not fatal to the X server /tmp/server-2.xkm

This is xvfb stderr leaking somehow to the xkbcomp process???

[Philippe Normand]

I'm trying an xorg update. I wonder if the old version, which we haven't bumped in 21.08 might have become incompatible with the xkbcomp that was updated in the parent SDK.

[Philippe Normand]

(In reply to Philippe Normand from comment #12)
> I'm trying an xorg update. I wonder if the old version, which we haven't
> bumped in 21.08 might have become incompatible with the xkbcomp that was
> updated in the parent SDK.

Issue still present after locally updating to xorg-server-1.20.13.

[Carlos Garcia Campos]

(In reply to Philippe Normand from comment #10)
> (In reply to Philippe Normand from comment #9)
> > There is an issue with the GTK API tests, possibly related with xvfb... I'll
> > check.
> 
> Yes... Carlos, there is a weird issue again with xvfb / subprocess.
> 
> If I redirect the stderr process to /dev/null the test runner is running OK.
> Otherwise it runs a few tests and hangs forever. My WIP diff is
> http://sprunge.us/3gJMX6
> 
> The SDK runtime has been updated from python 3.8 to python 3.9. Could this
> be related?

I'm having similar issues with the WebDriver tests under xvfb (no flatpak).

[Philippe Normand]

Applying that pastebin helps?

[Philippe Normand]

FTR this xvfb issue was worked-around in https://bugs.webkit.org/show_bug.cgi?id=230610

[Philippe Normand]

Committed r282878 (242007@main): <https://commits.webkit.org/242007@main>