View Issue Details
| ID | Project | Category | View Status | Date Submitted | Last Update |
|---|---|---|---|---|---|
| 0003669 | OpenFOAM | Bug | public | 2021-04-29 15:33 | 2021-04-30 12:23 |
| Reporter | tniemi | Assigned To | henry | ||
| Priority | low | Severity | minor | Reproducibility | sometimes |
| Status | resolved | Resolution | fixed | ||
| Product Version | dev | ||||
| Fixed in Version | dev | ||||
| Summary | 0003669: Regression in Time.C, first write time is wrong if not starting from 0 and not a restart | ||||
| Description | Hi We stumbled to odd behavior of sometimes not getting time directories as output if the starting folder was not 0 and it was not a normal restart, ie. no uniform/time file present which would contain the original beginTime. Managed to bisect it to this commit: https://github.com/OpenFOAM/OpenFOAM-dev/commit/bae95b1291c14fbc408a6f5533028ee8ab61f527 I think the problem is that beginTime_ is first initialized to 0 and setWriteInterval is called in readDict using this beginTime. Later on beginTime is set properly as beginTime_ = timeDict.lookupOrDefault("beginTime", startTime_); to be equal to non-zero startTime. But after this restart() is not true, so the writeTimeIndex_ is not updated and the invalid value set in readDict remains. Thus the first write will be "value()-0", which can be quite a big value if value is large. | ||||
| Steps To Reproduce | Take cavity tutorial, change: startTime 0->2 (and rename the 0-folder to 2) endTime 0.5->5 writeControl timeStep->runTime- writeInterval 20->0.1 First write will be 4.1, not 2.1 as it should | ||||
| Tags | No tags attached. | ||||
|
|
Shall I revert the commit? |
|
|
No, I think the commit is mostly fine. It is just that the logic regarding the very first writeTimeIndex_ calculation is slightly changed which breaks this one type of startup. I can perhaps provide a patch, but it is too late for today. It should be enough to just always update writeTimeIndex_ with beginTime=startTime if no other beginTime was found. I wonder is for instance the check for restart() in line 272 necessary? If the check would be removed, I think it would fix this issue. But not sure if it would break something else. |
|
|
Resolved by commit 6d7703bfa3467767faa3202a2d7b07cfd6bffde4 |
| Date Modified | Username | Field | Change |
|---|---|---|---|
| 2021-04-29 15:33 | tniemi | New Issue | |
| 2021-04-29 15:48 | henry | Note Added: 0012009 | |
| 2021-04-29 16:58 | tniemi | Note Added: 0012010 | |
| 2021-04-30 12:23 | henry | Assigned To | => henry |
| 2021-04-30 12:23 | henry | Status | new => resolved |
| 2021-04-30 12:23 | henry | Resolution | open => fixed |
| 2021-04-30 12:23 | henry | Fixed in Version | => dev |
| 2021-04-30 12:23 | henry | Note Added: 0012011 |
