avoiding memory leaks on errors in weekly schedule for #1860

[Mathadon]

@mwetter this should fix the memory leak issue. I was able to run IBPSA.Utilities.IO.Files.Examples.WeeklySchedule using this code but I did no memory checks using valgrind.

I don't have a machine to test the windows line endings. Is that something you could have a look at?

[mwetter]

Note that IBPSA.Utilities.IO.Files.Examples.WeeklySchedule crashes on my Dymola, with an empty log file.

[mwetter]

I could test the windows line endings, but can you please update the code. I am not familiar enough with the logic of your C implementation, by maybe where you test for \n, you could check if there is a \r followed by an \n and if so, skip the \r. Or alternatively convert the buffer from \r\n to \n before all the processing starts?

We should then replicate one of the validation and use a text file with Windows line endings, which we should be able to add with dos2unix.

[Mathadon]

I have made some revisions but can't figure out from the code why it crashes in Dymola. On OM it works ok:

I'm not sure how to debug further since we do not have a dymola license. You also checked the simulation tab for any error messages?

[Mathadon]

@mwetter through 0ecdc2e I attempted to add \r\n support.

[mwetter]

@Mathadon : I have not checked for the root cause of the error, but also OCT refuses the translate the file. It issues:

free(): double free detected in tcache 2

[Mathadon]

@mwetter thanks for checking further! I got the same error using Taco and apparently I forgot to remove the original fclose(). See ac9c6e2 for a fix. Does that resolve the windows issue too?

[mwetter]

@Mathadon : Now the results are wrong:

[Mathadon]

@mwetter thanks for the windows example model. I found the bug and tested the new example. The results visually match the old results in your screenshot and the new example works. :) see a911c6a

[mwetter]

@Mathadon : Thanks, this looks good to merge when the CI tests pass.

[Mathadon]

🙌