RE: [cdt-debug-dev] CValue.hasVariables()

[John Cortell <john.cortell@xxxxxxxxxxxxx>]

Hm. Ok, so your experience hints that the backend may not have the 
data during some initial window of queries, but once it does have it, 
CDT can then cache that information. Since to the caller, "I have no 
children" and "Info not available, so no children" are 
indistinguishable I guess the only recourse is to keep calling the 
CDI client until a non-zero/empty value is returned, if that ever 
happens. This may seem trivial, but this inconsistency can produce 
more than subtle (perhaps negligible) performance irregularities. I 
got on this path because of a quirk in our debugger where after 
changing memory via the Memory view, one variable in the Variables 
view updated but the same did not happen for another. The reason 
turned out to be that we're hitting the backend in one case but not 
the other, and that is ultimately due to the cache inconsistencies 
I've brought up.

I'll go ahead and open bugzilla reports for this stuff. Thanks for the insight.

John

At 05:52 AM 11/22/2006, Mikhail Khodjaiants wrote:
> John,
>
> I think it's an oversight. It wouldn't make much difference for the
> gdb/mi implementation, because the CDI values are also cached in the
> backend.
> Regarding your second question, I am not so sure. As I recall the first
> implementation was based on the logic you described, but there have been
> cases when it didn't work.
> In my opinion the variables related code is too complicated because of
> the many last minute changes. I wrote it, but now I think all caching
> should have been done in the backend.
>
> Mikhail
>
> -----Original Message-----
> From: cdt-debug-dev-bounces@xxxxxxxxxxx
> [mailto:cdt-debug-dev-bounces@xxxxxxxxxxx] On Behalf Of John Cortell
> Sent: 22 November 2006 04:18
> To: CDT Debug developers list
> Subject: [cdt-debug-dev] CValue.hasVariables()
>
> I found that CValue fails to use its cached information in its
> implementation of IValue.hasVariables(), resulting in a trip to the
> backend on every invocation. If we use a cache to avoid repeated trips
> to the backed for IValue.getValueString() and IValue.geVariables(), it
> seems to me that not doing so for
> hasVariables() is probably an oversight, and an opportunity for a
> performance improvement (could help make a heavily populated variables
> view a little snappier; this method gets called fairly often).
>
> On the same token, the cache logic in getVariables0() seems a bit
> flawed. Effectively, it says: if our cached children list is empty, then
> we must have not gotten the list from the backend, so try to get it now.
> But in fact, we might have been there before--already asked the backend
> and it told us the variable has no children. So, really, we shouldn't be
> asking again. Not sure how much we stand to gain here, but if we're
> going to use a cache in CValue, we should probably use it in a
> consistent manner.
>
> I wanted to pose the question before opening a bugzilla report, in case
> I'm overlooking something.
>
> John
>
> _______________________________________________
> cdt-debug-dev mailing list
> cdt-debug-dev@xxxxxxxxxxx
> https://dev.eclipse.org/mailman/listinfo/cdt-debug-dev
>
> --
> IMPORTANT NOTICE: The contents of this email and any attachments are 
> confidential and may also be privileged. If you are not the intended 
> recipient, please notify the sender immediately and do not disclose 
> the contents to any other person, use it for any purpose, or store 
> or copy the information in any medium.  Thank you.
>
>
> _______________________________________________
> cdt-debug-dev mailing list
> cdt-debug-dev@xxxxxxxxxxx
> https://dev.eclipse.org/mailman/listinfo/cdt-debug-dev