Build 958 Crash

[Synewave]

You could also look in the diags log to see what errors were logged.

The diags, the only error I can see is "Yesterday file is valid".

I've looked at yesterday.ini and can't find any errors.

[steve]

I've looked at the code and I can't explain how this month and this year can get different values and/or times when they both occur this month. I must be missing something. Unless there were indeed errors processing the data which affected the setting up of the monthly and yearly data differently.

[steve]

the only error I can see is "Yesterday file is valid".

That doesn't look like an error to me

[steve]

I should be able to reproduce your problem quite easily; if you let me have your dayfile.txt and today.ini as they were at the time you first ran this build - take them from the backups folder.

[Synewave]

Re: Yesterday file is valid, why did I read that as invalid? After helping Antonio yesterday I think I've damaged my brain!

Anyway, sorry for the false alarm, here are the files:

[steve]

I get this error in the diags file:

02/11/10 12:44:03.595 : Error reading Dayfile around line 93: '' is not a valid floating point value

01/01/10,6.9,315,00:31,-0.3,23:31,6.2,13:01,998.6,01:01,1011.3,23:31,0.0,00:01,0.0,2.5,29.5,,

It's the null entry at the end; all of your entries from the start of the file up to and including 01/02/10 have the same thing. I have no idea how that's happened, and there's no code in Cumulus to cope with that.

[Synewave]

Interesting, I've just looked at my backup of the dayfile from prior to 954 and it is the same, so it's definately not builds after 954 that may have caused a problem.

[Synewave]

There also seems to be a pattern with the strange records on my page, times are 01:01 on 01 January and 01:11 on 01 November.

[steve]

Interesting, I've just looked at my backup of the dayfile from prior to 954 and it is the same, so it's definately not builds after 954 that may have caused a problem.

Cumulus doesn't put those blank fields in - I don't have them in my file. And it never rewrites the entries, so once an entry is in there, a subsequent build of Cumulus won't change them.

There's very little validation of the format of the log files in Cumulus; it never occurred to me to put any in, as only Cumulus writes the files so it knows the format should be OK. I guess that now people are having to edit files because of bad data, I ought to put more checking in. This new code does actually check for and ignore blank lines!