Hi Devs, This is not really a "bug-report", but I want to mention it nonetheless, since engine-crashes are probably not what should be left unhandled, regardless if the user tries to use the engine in an undefined-mode.
Let me explain: Background: Platform Windows. StdCall-compile of the original "amalgamation". COM-Wrapper "on top". Crashes with each of the newer versions (3.6.9 and 3.6.10) - so, yes - the whole thing is probably related to the changes in the internal transaction- support of the engine. A user reported "full crashes" in a larger "copy- over-scenario" from another DB (FireBird). Switching back to the former engine 3.6.7 was working Ok. Finally the cause was a "left-over" switch, somewhere hidden in an Ini-File, which was introduced regarding "performance tuning" (together with switches for the different SyncModes and different PageSizes). This forgotten switch was causing a JournalMode = Off. Later on Transactions were used inside the Copy-over- process from FireBird. And that is, what I mean with "undefined mode" - Journal-Mode=Off and nonetheless "transactions on top" were somehow tolerated by the older engines < 3.6.9 - but the newer engines are not tolerating that anymore (and cause a crash). The transactions in question were used with the "older" standard-syntax (Begin Immediate, etc.) and not with the new introduced SavePointName- syntax. Regards, Olaf _______________________________________________ sqlite-users mailing list sqlite-users@sqlite.org http://sqlite.org:8080/cgi-bin/mailman/listinfo/sqlite-users
