Spam logs: "Could not find an exact match rule. Using 'closest' rule"

[Colin Alworth]

It looks like changes to a few rebind rules are generating some new logspam when any GWT app compiles. Specifically, I'm seeing FocusImpl and LayoutImpl, though its possible there are others I haven't seen yet. From the dynatable example we can see the FocusImpl spam:
gwtc:
     [java] Compiling module com.google.gwt.sample.dynatable.DynaTable
     [java]    Computing all possible rebind results for 'com.google.gwt.user.client.ui.impl.FocusImpl'
     [java]       Rebinding com.google.gwt.user.client.ui.impl.FocusImpl
     [java]          Could not find an exact match rule. Using 'closest' rule <replace-with class='com.google.gwt.user.client.ui.impl.FocusImplIE6'/> based on fall back values. You may need to implement a specific binding in case the fall back behavior does not replace the missing binding
     [java]    Compiling 5 permutations
     [java]       Compiling permutation 0...
     [java]       Process output
     [java]          Compiling
     [java]             Compiling permutation 1...
     [java]       Compiling permutation 2...
     [java]       Compiling permutation 3...


First, the rules themselves:
  <!-- Firefox uses a hidden input to set accesskeys -->
  <replace-with class="com.google.gwt.user.client.ui.impl.FocusImplStandard">
    <when-type-is class="com.google.gwt.user.client.ui.impl.FocusImpl"/>
    <when-property-is name="user.agent" value="gecko1_8"/>
  </replace-with>

  <!-- Safari uses a hidden input to set accesskeys and -->
  <!-- fires focus/blur after a timeout                 -->
  <replace-with class="com.google.gwt.user.client.ui.impl.FocusImplSafari">
    <when-type-is class="com.google.gwt.user.client.ui.impl.FocusImpl"/>
    <when-property-is name="user.agent" value="safari"/>
  </replace-with>

  <!-- IE's implementation traps exceptions on invalid setFocus() -->
  <replace-with class="com.google.gwt.user.client.ui.impl.FocusImplIE6">
    <when-type-is class="com.google.gwt.user.client.ui.impl.FocusImpl"/>
    <all>
      <any>
        <when-property-is name="user.agent" value="ie6"/>
        <when-property-is name="user.agent" value="ie8"/>
      </any>
      <none>
        <when-property-is name="user.agent" value="ie9"/>
      </none>
    </all>
  </replace-with>

No express binding of FocusImpl to itself, or to Standard (interestingly Standard really seems to mean Gecko, at least according to that comment, calling out a FF specific issue). I can't comment much on that just yet, though I might make a suggestion later.

This code is as a result of https://gwt-review.googlesource.com/#/c/5055/ (or https://gwt.googlesource.com/gwt/+/040b3e4392186e48689a6fc1f19cdf294f2b5651 if you like), ostensibly to make IE9+ use FocusImpl instead of FocusImplIE6 (or Standard, for example).

[java]    Computing all possible rebind results for 'com.google.gwt.user.client.ui.impl.FocusImpl'

...

[java]          Found better fallback match for <replace-with class='com.google.gwt.user.client.ui.impl.FormPanelImpl'/>
[java]          Checking rule <replace-with class='com.google.gwt.user.client.ui.impl.FocusImplIE6'/>
[java]             Checking if all subconditions are true (<all>)
[java]                <when-type-is class='com.google.gwt.user.client.ui.impl.FocusImpl'/>
[java]                   Yes, the requested type was an exact match
[java]                Checking if all subconditions are true (<all>)
[java]                   Checking if any subcondition is true (<any>)
[java]                      <when-property-is name='user.agent' value='ie8'/>
[java]                         Property value is 'ie9'
[java]                         Property value 'ie8' is the fallback of '[[ie9]]'
[java]                         No, the value did not match
[java]                      No: All subconditions were false
[java]                   No: One or more subconditions was false
[java]                No: One or more subconditions was false
[java]          Rule did not match
[java]          Found better fallback match for <replace-with class='com.google.gwt.user.client.ui.impl.FocusImplIE6'/>
[java]          Checking rule <replace-with class='com.google.gwt.user.client.ui.impl.FocusImplSafari'/>
[java]             Checking if all subconditions are true (<all>)
[java]                <when-type-is class='com.google.gwt.user.client.ui.impl.FocusImpl'/>
[java]                   Yes, the requested type was an exact match
[java]                <when-property-is name='user.agent' value='safari'/>
[java]                   Property value is 'ie9'
[java]                   No, the value did not match
[java]                No: One or more subconditions was false
[java]          Rule did not match
[java]          Checking rule <replace-with class='com.google.gwt.user.client.ui.impl.FocusImplStandard'/>
[java]             Checking if all subconditions are true (<all>)
[java]                <when-type-is class='com.google.gwt.user.client.ui.impl.FocusImpl'/>
[java]                   Yes, the requested type was an exact match
[java]                <when-property-is name='user.agent' value='gecko1_8'/>
[java]                   Property value is 'ie9'
[java]                   No, the value did not match
[java]                No: One or more subconditions was false
[java]          Rule did not match

...

[java]          Could not find an exact match rule. Using 'closest' rule <replace-with class='com.google.gwt.user.client.ui.impl.FocusImplIE6'/> based on fall back values. You may need to implement a specific binding in case the fall back behavior does not replace the missing binding
[java]          No exact match was found, using closest match rule <replace-with class='com.google.gwt.user.client.ui.impl.FocusImplIE6'/>
[java]          Rebind result was com.google.gwt.user.client.ui.impl.FocusImplIE6

From this abbreviated log, apparently IE9 is still using the IE6 impl, rather than ignoring any rebind rule and using the actual argument to GWT.create.

My *guess* is that this has been the case since the patch linked was authored, and checking out 040b3e4392186e48689a6fc1f19cdf294f2b5651 seems to confirm this - building dynatable still has this log message, indicating that IE9 has been using IE6 all along, despite the commit message. The linked issue https://code.google.com/p/google-web-toolkit/issues/detail?id=8383 (which I didn't read until just now, and could have saved me a fair bit of time...) seems to confirm this.

My goal in this quick bit of debugging is to get us to a stable GWT release that doesn't add useless logging. To that end, I think 040b3e4392186e48689a6fc1f19cdf294f2b5651 must be reverted, since it has no effect, and adds confusing log messages that might suggest to the user that they have made a mistake. The commit at https://gwt.googlesource.com/gwt/+/a0a0c20713cbc2d0fc8dd262f066da9f2412c93a should get the same treatment (LayoutImpl), and I'll be walking a few more commits to confirm that everything else looks safe.

The other option is to declare explicitly to bind FocusImpl to FocusImpl (with or without a 'useragent=ie9' check) so that this is then the closest matching rule. Since we haven't actually been testing IE9 with the 'real' implementation all along, we have no real idea right now what would happen if we did this, so I'm inclined against it.

Thoughts or concerns about this issue and any fix?

[Goktug Gokdogan]

Reverting the binding sounds good to me. Please add a comment to the binding; it might probably be bound to something else but not tested.

--
You received this message because you are subscribed to the Google Groups "GWT Contributors" group.
To unsubscribe from this group and stop receiving emails from it, send an email to google-web-toolkit-co...@googlegroups.com.
To view this discussion on the web visit https://groups.google.com/d/msgid/google-web-toolkit-contributors/e1628118-bb35-4d5a-aa31-b9ba1ce070ba%40googlegroups.com.
For more options, visit https://groups.google.com/d/optout.

[Colin Alworth]

Sorry Goktug, I don't follow - what should the comment indicate? In the old code, IE9 was expressly bound to the IE6 imp, in the new (to be reverted) code, it was expressly *not* bound to the IE6 impl, but it was the closest fallback anyway.

Are you suggesting that the comment should point out that IE9 probably shouldn't be bound in that way (though it works, and may not work if differently bound)?

Thanks for any clarification,
Colin

To view this discussion on the web visit https://groups.google.com/d/msgid/google-web-toolkit-contributors/CAN%3DyUA26VSXTzM9MviU34o80e%3D-NCchT0qdkPu3MSeW04epceA%40mail.gmail.com.

For more options, visit https://groups.google.com/d/optout.

--
218.248.6165
nilo...@gmail.com

[Goktug Gokdogan]

The comment is just to indicate that IE9 is bound to legacy FocusImplIE6 only because we didn't test is with a more modern implementation. So later when somebody looks at the code for cleanup will not think that we carefully tested and decided to use the Iegacy implementation.

To view this discussion on the web visit https://groups.google.com/d/msgid/google-web-toolkit-contributors/CADcXZMwNNrepDuCuuopGg1rhboctaYvd7mGjFY69o00oPw9GBA%40mail.gmail.com.

[Jens]

Maybe someone wants to mark one of the issues as duplicated since some time ago I also created an issue for it as I haven't seen the one that Colin has found.

However in our compiles it is not only FocusImpl and LayoutImpl but also InternalJsHashCodeMap and InternalJsStringMap, see:

https://code.google.com/p/google-web-toolkit/issues/detail?id=8848

-- J.