Scheduled transaction issue

[John Ralls]

On Apr 30, 2012, at 7:18 PM, Donald Allen wrote:

> On Mon, Apr 30, 2012 at 5:25 PM, Donald Allen <donaldcallen at gmail.com> wrote:
>> On Mon, Apr 30, 2012 at 3:57 PM, John Ralls <jralls at ceridwen.us> wrote:
>>> 
>>> On Apr 30, 2012, at 8:20 AM, Donald Allen wrote:
>>> 
>>>> My gnucash data is sitting in postgres on a Linux system. I normally
>>>> access it by running gnucash (2.4.10) on the same system. I have a
>>>> number of scheduled transactions that are set up to happen near the
>>>> end of every month. A number of those transactions got recorded
>>>> yesterday, as a result of my running gnucash.
>>>> 
>>>> My wife runs Windows 7 on her laptop, and I've installed gnucash
>>>> 2.4.10 on her machine. I've pointed it at the postgres server on the
>>>> Linux machine mentioned above. This setup is new (I had previously
>>>> been working with .xml files). She and I were just working on her
>>>> machine, looking at some financial matters with gnucash. But when I
>>>> started gnucash on her machine, it offered to enter the same scheduled
>>>> transactions that had happened yesterday when I ran gnucash on my
>>>> Linux box. I declined the offer, of course, and then verified that I
>>>> was seeing the transactions that had been entered yesterday on the
>>>> Linux system. They were there. This feels like a bug, a serious one.
>>>> 
>>>> Thinking about what might cause this incorrect behavior makes me
>>>> wonder where the information about what scheduled transactions have
>>>> been run is stored. Is it stored in the database itself, as I would
>>>> expect, since the fact that the transactions have been entered is a
>>>> property of the data, not of a particular user or gnucash instance. Or
>>>> is it stored in individual home directories in, say, ~/.gnucash
>>>> directories (or whatever the analogous place is in the Windows world)?
>>>> If the latter, that would explain this behavior, but there are other
>>>> possible explanations, so I'm indulging in speculation here. So I'll
>>>> stop and turn this question over to people who actually understand the
>>>> code.
>>> 
>>> Odd indeed. The last-run date is stored in a normal table in the database and on my main accounts (kept in SQLite), it appears to be posted correctly.
>>> 
>>> Does "select name, last_occur from schedxactions;" produce sensible results?
>> 
>> Yes. Two transactions that it offered to enter today were entered on
>> 2012-04-28. Bizarre. I just tried it on another Windows 7 laptop and
>> got the same result as I did with my wife's machine -- it asked me to
>> allow it to enter transactions that were entered on 4/28. This is is a
>> bug. I will try this with a Gnucash instance running on Linux, but
>> pointed at the same database server as the Windows machines, to see if
>> this is Windows-specific.
> 
> I just repeated the experiment on the same laptop I referred to in the
> above paragraph (it's setup dual-boot), and running gnucash 2.4.10
> pointed at the same remote database on Linux does *not* produce the
> symptom I reported, i.e., it does not incorrectly offer to enter
> scheduled transactions that have already been entered. This appears to
> be a Windows-specific problem.
> 
> Any ideas?

Interesting. That's two M$Win-only  SQL problems in a week. That's a major PITA just because debugging in M$Win is 10 times harder than on unix. 

No, no ideas at the moment. You'd better file a bug so that I don't forget about it.

Regards,
John Ralls