[QBS] Performance regression with latest QtCreator (3.2 branch)

[Christian Kandeler]

On 07/24/2014 04:03 PM, Thomas Epting wrote:
> Hi everyone,
>
> It seems there is a notable performance regression when using the latest
> QtCreator / Qbs from the 3.2 branch.
>
> I'm doing these actions for the same (non-trivial) project in 3.1.2 and
> 3.2 Creator:
>   1. Do a „null“ build (nothing to do) several times, use “Elapsed time”
> from last try.
>   2. Select Build | Reparse Qbs several times, clock each time and use
> average.
>   3. Add a newline in any .qbs file and press save. Retry several times,
> clock until “Evaluating” becomes green and use average.
>
> Timings for QtCreator 3.2:
>   1. Null Build: 0:27
>   2. Reparse Qbs: 0:01
>   3. Evaluation after null-change + Save: 0:36
>
> Timings for QtCreator 3.1.2:
>   1. Null Build: 0:03
>   2. Reparse Qbs: 0:01
>   3. Evaluation after null-change + Save: 0:09
>
> Any ideas why everything takes so much longer now?

Creator now evaluates the project before building, because that is 
necessary to ensure build graph consistency (see QBS-596). On the other 
hand, the unnecessary reparsing step after a build was not necessary and 
has gone entirely. My experience is that there's way less time spent on 
parsing now.


Christian