Re: CellTable Memory Leak in IE8

No worries, glad it worked! ;)

I've not personally committed any code to the core gwt project but you can find out more here http://www.gwtproject.org/makinggwtbetter.html#contributingcode

Bear in mind that GWT 3.0 is going to be a very different animal (for instance, from what I understand there will be no widgets or gwt-rpc...), so perhaps there isn't such a need to get these things fixed... I might be wrong though.

Cheers,

Dave

On Wednesday, 5 August 2015 13:24:27 UTC+1, Harvard Pan wrote:

Dave,

Thank you SO much for this. I tested this using my test project and it seems to have solved the memory leak completely! I've attached my version of the file (modified from 2.5.1-rc1 branch) with your changes.

What's the best way to get this fix to the GWT code base so that future versions of GWT will have this fix as well? I'd previously logged Issue #9164 (https://github.com/gwtproject/gwt/issues/9164) on the gwtproject page.

Thanks again!

Harvard

On Tuesday, August 4, 2015 at 7:11:24 AM UTC-4, DaveC wrote:

Hi,

I think you'll have to dig into AbstractCellTable in order to fix this. I did a quick test of a fix I implemented and it "appears" to work for IE11 running IE8/IE9 mode but not IE10 mode, I've not tested it in a real IE8 or IE9.

Basically what I did was exaactly what you said - set row.innerHTML to null.

I took a copy of AbstractCellTable (including the package structure) and placed it in my test project. In the ImplTrident inner class I've tweaked the methods replaceTableSection and replaceAllRowsImplLegacy to set the innerHTML to null for each row e.g. 

    private native final void setRowInnerHtmlToNull(Element row)/*-{

    row.innerHTML = null;

    }-*/;

private void replaceTableSection.....

 TableElement tableElement = table.getElement().cast();

      

      Element child = section.getFirstChildElement();

      while (child != null) {

        setRowInnerHtmlToNull(child);

        child = child.getNextSiblingElement();

      }

      

      tableElement.replaceChild(newSection, section);

.....}

 protected void replaceAllRowsImplLegacy(AbstractCellTable<?> table, TableSectionElement section,

        SafeHtml html) {

      // Remove all children.

      Element child = section.getFirstChildElement();

      while (child != null) {

        Element next = child.getNextSiblingElement();

        section.removeChild(child);

        setRowInnerHtmlToNull(child);

        child = next;

      }

....}

Hope this helps.

Cheers,
Dave

On Thursday, 23 July 2015 22:05:10 UTC+1, Harvard Pan wrote:

Hello,

Our company uses GWT 2.5.1-rc1 and many of our customers (healthcare) use IE8. We were hopeful that the memory leak in CellTable would have been addressed by the memory leak fix for FlexTable. That leak (Issue 6938 - https://code.google.com/p/google-web-toolkit/issues/detail?id=6938) was fixed in 2.6. After grabbing the fix and merging it into the 2.5.1-rc1 code, we can confirm that FlexTable indeed is fixed and no longer leaking. However, we still have leaking resources in CellTable. I've written a small sample application to demonstrate the code that leaks. It's available on BitBucket for anyone to pull.

https://bitbucket.org/harvardpan/celltableleak/src/3eb0d9941df6fe40b4b09eef0ce1968c6db90da3/src/com/healthfortis/sample/celltableleak/client/CellTableLeak.java?at=master

The root cause of 6938 was described in a Microsoft Connect page: http://connect.microsoft.com/IE/feedback/details/790340/memory-leak-in-ie9-ie10-tables

In it, it describes multiple reasons for the leak, including:

• rows with ids
• cells with ids
• code that references a row.cells expression, even if it does not store or use the result (that's the fix in 6938)
• code that does not set row.innerHTML to null after invoking table.deleteRow() for the row.

I imagine that the CellTable leak is related to one of the conditions above. I suspect the last one as I never do actually see any setting to null of innerHTML in the javascript. Wanted to check in to see if anyone on this forum had any ideas on where we could investigate next.

Thanks!

Harvard

--
You received this message because you are subscribed to the Google Groups "Google Web Toolkit" group.
To unsubscribe from this group and stop receiving emails from it, send an email to google-web-toolkit+unsubscribe@googlegroups.com.
To post to this group, send email to google-web-toolkit@googlegroups.com.
Visit this group at http://groups.google.com/group/google-web-toolkit.
For more options, visit https://groups.google.com/d/optout.