Comment 1 Nobody - feel free to take it 2015-09-09 16:36:00 EDT

Thank you Dawid. This actually happens on PDT 3.5, and is visibly (in part?) a regression from PDT 3.4.

To clarify, this is issue #2 discussed in the forum message. When the Variables view is refreshing, its title is in italics.

The refresh time depends on the project. I noticed this issue in a proprietary project where in certain areas (I believe only in the global scope), any statement (even the fastest) would take at least 1 or 2 seconds to step over. This minimum time taken clearly depends on the variables in the scope.

Unfortunately, the code is huge and I have no rights over it, so I cannot share. It would be interesting to produce a test case where stepping is slow and see if others reproduce, but I am not sure how to do that. Would simply declaring a huge string cause it, or does it depend on the number of variables (and their types)?

Comment 2 Nobody - feel free to take it 2015-09-09 16:40:42 EDT

Bartlomiej Laczkowski has asked for my .log file. I hesitate to attach it since it is full of errors (almost 1 MB). But there are no errors generated when this "problem" appears. The last errors logged are from before:

!ENTRY org.eclipse.php.debug.core 4 4 2015-09-09 15:00:18.069
!MESSAGE class org.eclipse.php.internal.debug.core.xdebug.dbgp.model.DBGpTarget : DBGp Response Error: eval:=206 msg:error evaluating code

!ENTRY org.eclipse.php.debug.core 4 4 2015-09-09 15:00:21.507
!MESSAGE class org.eclipse.php.internal.debug.core.xdebug.dbgp.model.DBGpTarget : DBGp Response Error: eval:=206 msg:error evaluating code

Comment 3 Bartlomiej Laczkowski 2015-09-09 17:21:33 EDT

Hello Filipus,

Did you try to change the default value of "Max children" option to eg. 20 in XDebug preferences (Window -> Preferences -> PHP -> Debug -> Debuggers)? Can you try to change it and see if that speeds up refreshing of Variables view?

Comment 4 Nobody - feel free to take it 2015-09-10 00:14:05 EDT

Sorry Bart, I had forgotten to.

The section I am currently using to reproduce has each step taking about 1 second. It is clear that changing from 51 to 20 does not cause steps to take 20/51 of a second. I do not see an obvious difference, but it is non-trivial to time a second precisely, so I cannot say this causes no significant change. If anyone has a simple tip to slow down the CPU by an order of magnitude on Windows 8.1, I can retest.