FWD » Core Development » User Interface

In OE CHARACTER mode, pressing Enter produces no visible effect, whereas in FWD, it does. Thus, the same issue occurs in ChUI mode as well.

The example used in both modes is:

SESSION:DATE-FORMAT = 'mdy'.
DEFINE VARIABLE dateField AS DATE VIEW-AS FILL-IN NO-UNDO FORMAT '99/99/9999'.
DEFINE FRAME f dateField NO-LABEL.
ENABLE ALL WITH FRAME f.
WAIT-FOR WINDOW-CLOSE OF FRAME f.

The issue originates in the following code from FillIn.processKeyEvent() (line 1737):

  if (key == Keyboard.KA_RETURN)
  {
     if (isLinkedToBrowse())
     {
        ...
        return;
     }

     if (!completeEdit())
     {
        return;
     }

     if (ThinClient.getInstance().isChui() && isLastWidget)
     {
        ... do something
     }
     else if (getDataEntryReturn() || (ThinClient.isDataEntryReturn() && ke.isTriggerFired()))
     {
        if (groupMode && !isLastInGroup())
        {
           ... do something
        }
        else
        {
           if (isLastWidget)
           {
              ... do something
           }
           else
           {
              ... do something
           }
        }
     }

     refresh();
     return;
  }

The problem is that completeEdit() is called unconditionally even if no action for the RETURN key is selected. To resolve this:

1. Identify which actions (marked as "... do something") require a prior call to completeEdit().
2. Remove the unconditional completeEdit() call and add it only in the branches that truly require it.

The Fix¶

The revised code relocates the completeEdit() call to the start of the relevant branch and moves refresh() to the end of that branch as well. The ThinClient.getInstance().isChui() && isLastWidget branch, which simply posts the GO event in CHARACTER mode, requires neither:

  if (key == Keyboard.KA_RETURN)
  {
     if (isLinkedToBrowse())
     {
        ... do something
        return;
     }
     ...
     if (ThinClient.getInstance().isChui() && isLastWidget)
     {
        ... post GO event
        return;
     }

     if (getDataEntryReturn() || (ThinClient.isDataEntryReturn() && ke.isTriggerFired()))
     {
        // FIXME: Do we need this here?
        if (!completeEdit())
        {
           return;
        }

        if (groupMode && !isLastInGroup())
        {
           ... do something
        }
        else if (isLastWidget)
        {
           ... do something
        }
        else
        {
           ... do something
        }

        refresh();
     }

     return;
  }

Notes on the Changes:

1. The getDataEntryReturn() || (ThinClient.isDataEntryReturn() && ke.isTriggerFired()) branch has not been fully analyzed. Its logic appears complex and may introduce other discrepancies, but this fix should not affect its behavior or cause regressions.

Code review 10876a.

completeEdit is now not called for last widget in ChUI mode. Is this expected?

Why was the FIXME comment added, does it refer to completeEdit or the return statement?

Hynek Cihlar wrote:

Code review 10876a.

completeEdit is now not called for last widget in ChUI mode. Is this expected?

Yes. See #10976-2 above:

In OE CHARACTER mode, pressing Enter produces no visible effect, whereas in FWD, it does. Thus, the same issue occurs in ChUI mode as well.

Why was the FIXME comment added, does it refer to completeEdit or the return statement?

My change does not affect this code fragment. This comment can be removed.

Hynek Cihlar wrote:

Vladimir, please address the FIXME comment and go ahead with regression testing. ChUI regression tests will be also needed. Are you able to run them already?

Done in 10876a rev. 16332.

The fix also resolves #9913 specifically for the Enter key.

I verified it using the example from #10876-2 in both GUI and ChUI modes.
Additionally, I tested it with our large customer application - everything remains stable.

Hynek, run the ChUI test suite on my behalf.

ChUI tests running.

Vladimir, I found some regressions when running ChUI regression tests. Please see the results in devsrv01:/tmp files results_20260204132130.zip and results_20260204091019.zip, one set of results is from trunk the other from the branch.

Vladimir, please make sure to check in any test cases that are still unchecked.

Hynek Cihlar wrote:

Vladimir, please make sure to check in any test cases that are still unchecked.

No test program other than the one in #10876-2 is required.

10876a rebased to r16411 and retested - no issues found.