Why is "Show uses" sometimes so slow?

[Alexandre Courbot]

Hi Andreas,

On Mon, May 28, 2012 at 4:28 AM, Andreas Pakulat <apaku at gmx.de> wrote:
> The parsing of theproject sources only retrieves some of the available
> information. Mostly whats needed to supply auto-completion and type
> information. Calculating uses is only done on demand (and then cached),
> since it requires an extraordinary amount of time to be done for all code
> and also increases the memor footprint significantly. And its not necesary
> for many of the smbols in a project.
>
> Personally I do mind if code-completion and other things take considerable
> time after opening a project the first time, since it means kdevelop is less
> useful and slows down working.

Yes, my issue is that I will be giving a live demo of KDevelop at
LinuxCon Japan in a little bit more than one week on the kernel
sources, and although everything else is snappy "Show Uses" takes an
insane amount of time if I look for anything that is not local to a
couple of files. Later searches are instant, but for the purpose of
the demo I'd like to get the first one as fast as possible too.

> That being said, IIRC you can change the amount of information gathered in
> the kdevelop settings for the parser.

Not that I know - the parser settings currently only includes the
delay an number of threads. In language support you can choose the
level of code completion, but I don't think this is related to the
parser.

I will try to look at the parser/DUChain building before the event,
but if someone could teach me a quick hack that could speed things up
(I am not too concerned about initial parsing time or used disk space
- currently the initial DUChain for the kernel takes less space than
CTags generated information), I'd be most grateful.

Thanks,
Alex.