VisualD/cv2pdb doesn't display AA's properly in the watch window.

[Lewis via Digitalmars-d-debugger]

On Thursday, 18 February 2016 at 23:01:24 UTC, Rainer Schuetze 
wrote:
>
>
> On 18.02.2016 01:36, Lewis wrote:
>> Test code (in a newly created project):
>>
>>
>> import std.stdio;
>>
>> int main(string[] argv)
>> {
>>      writeln("Hello Blah!");
>>
>>      string[string] testAA = null;
>>      testAA["test1"] = "Hello";
>>      testAA["test2"] = "Goodbye";
>>      testAA["blah"] = "string";
>>      testAA["bloop"] = "another string";
>>      testAA["words"] = "test string";
>>
>>      int[] testArray = null;
>>      testArray ~= 3;
>>      testArray ~= 10;
>>      testArray ~= 11;
>>      testArray ~= -2;
>>      testArray ~= 14;
>>
>>      readln();  // Breakpoint on this line
>>      return 0;
>> }
>>
>>
>> When I check the watch window, I see the following:
>> http://imgur.com/mxOxitP
>>
>> I'm using DMD 2.069.0. The project is set to use the VS 
>> debugger and
>> cv2pdb. I'm using VS2015, with native compatibility and edit 
>> and
>> continue enabled: http://imgur.com/6mBDtN4
>>
>> The debugger displays the dynamic array just fine. It correctly
>> identifies the AA as an aa2, complete with types, and knows 
>> its size. It
>> just doesn't display its contents correctly. I've tried 
>> reinstalling
>> VisualD with no luck. I've checked the autoexp.dat and the 
>> VisualD
>> entries are there as expected. I tried mucking with 
>> autoexp.dat a bit to
>> see if I could make it work myself, but to no avail, although 
>> that tells
>> me VS is indeed using autoexp.dat as I'd expect.
>>
>> Any ideas what's going on? My suspicion is that something in 
>> D's
>> representation of AAs changes slightly, and now the 
>> autoexp.dat just
>> needs a tweak, but I don't know nearly enough to know what to 
>> fix myself.
>>
>> Thanks in advance!
>
> cv2pdb has not been updated to the new AA implementation in dmd 
> 2.068 
> (http://dlang.org/changelog/2.068.0.html#aa-open-addressing) 
> yet. That's why the type information generated by cv2pdb does 
> not fit and autoexp.dat cannot extract the info.
>
> IIRC mago has been updated, so you might want to use this 
> debugger engine instead.

Gotcha, thanks for the reply! Just tested with Mago, and it 
indeed shows AAs as expected.

As an aside, I noticed a quirk when viewing AAs with Mago in VS. 
If I have an AA of strings to structs, I can see all the keys and 
struct addresses no problem. However, I can't drill down further 
and inspect the contents of an individual struct. That being 
said, I can create a second watch of the form 'testAA["key"]', 
and now I can see the contents of that struct.

Just curious if there is a technical or design limitation 
preventing this, or if it just never got implemented. If it's the 
latter, may I humbly add this as a feature request for whenever 
yourself or another developer who knows Mago has time? I'm 
guessing I'd just add an issue on the github page?

Thanks!

Lewis