On 24 Oct 2015, at 8:23pm, Catalin Ionescu <catalin.ionescu at radioconsult.ro> wrote:
> So far I never used the command-line shell. It is clearly compiled with > different optimization settings and that might be the cause of apparent > correct functionality. Anyway, I decided to give it a try and attack the > problem as a whole (I suspected, from previous experience, that the bug is > just randomly appearing!). So I have fired up the shell, opened the DB file, > pasted the CREATE TEMP VIEW... statements and executed "select > sum(Quantity*PriceMed) from Stocks_View;". With 3.8.11 and newer I get a > results around 20528.817 and with 3.8.10.2 and older I get a results around > 11383.248. The value with the older SQLite versions is the correct one! > > Is this considered a satisfactory proof that there is a problem starting with > version 3.8.11? Maybe. It does at least give the development team something they can test without having to have a copy of your program. There are some ambiguities in the specifications for SQL and SQLite, where the documentation says that something is undefined or perhaps doesn't mention it at all. The thing you noticed may be one of them. They usually involve NULLs or other troublesome thing like that. Another possibility is that your database is corrupt. Find this out using <https://www.sqlite.org/pragma.html#pragma_integrity_check> There are two ways to proceed: A) put a copy of your database somewhere where the dev team can download it B) simplify your problem set by deleting tables and rows to the point where the commands to reproduce the problem are short. (You can obtain the commands needed using the .dump command for the command-line shell.) Then just post here listing the commands needed as text in your post. Simon.
