improved ill-posed integer rounding

[JayHuLBL]

This closes #2088.

[mwetter]

This still has potential rounding issues.

[mwetter]

This looks good now.

[mwetter]

@jelgerjansen : can you please provide a review so we have a non-LBL participant approving it. Thanks.

[jelgerjansen]

@JayHuLBL thanks for looking into this.

I noticed that you still applied some changes (commit 10f3e56) since Michael's approved review (commit 8462d07). Why did you split the initialization event, which is now still using the integer() implementation, from the runtime events, using the numerically more robust implementation?

Additionally, can you also:

1. Move the documentation (lines 23-24) to the correct part of the code (modTimAux > pre(tNext))?
2. Refer to the IBPSA issue next to the Buildings issue in the review notes (for completeness)?

[JayHuLBL]

@jelgerjansen The changes done in commit # 10f3e56 is to consider case when the simulation starts from the second half of the year.

For instance, if a simulation's start time is 190 days, then floor(modTimAux/lenWea + 0.5)*lenWea + lenWea will be 2*lenWea, which should be 1*lenWea.

Thanks for other comments. I will change it accordingly.

[mwetter]

@JayHuLBL : I still don't understand why there are two different equations to assign tNext.
If there is indeed a rational, it must be documented in the .mo file so users and developers understand.
Also, is the start attribute needed for tNext if it is assigned in an initial() clause?

[JayHuLBL]

The changes here is to address the potential issue that when the event is triggered due to the condition canRepeatWeatherFile and modTimAux > pre(tNext)) (thus for example modTimAux/lenWea = 1.00001), there is potential that it is triggered a little bit earlier (as described in (lbl-srg/modelica-buildings#4480) (thus for example modTimAux/lenWea = 0.99999). By changing the floor(modTimAux/lenWea) to floor(modTimAux/lenWea + 0.5) will ensure the right tNext calculation.

However, for the initial event, it is clear that it is triggered at the right time and it is safe to calculate tNext as floor(modTimAux/lenWea)*lenWea + lenWea.

[JayHuLBL]

@mwetter The implementation has been further improved through commit 4b45e33. Would you please review if the change looks good?

[mwetter]

See inline comment.

[mwetter]

Why is there an if-then-else needed? If we are here, then canRepeatWeatherFile must be true.

[JayHuLBL]

Changed in commit ecffbe9.