On Mon, Jun 21, 2010 at 9:37 AM, Roger Binns <rog...@rogerbinns.com> wrote:
> As an example you would experience something like this if SQLite got > initialized during the config phase. Just to make things even more > complicated ASLR is turned off if you have Wine installed under Linux so I > was only seeing the problem on some machines. Roger, Your post bring me back to what I asked in my last post: On Sun, Jun 20, 2010 at 11:31 PM, Sam Carleton <scarle...@miltonstreet.com> wrote: > > Through one request from the client, I open and close the database > multiple times. Since everything is very module, I am taking that > approach with the DB, too: For authentication it is open/closed, for > the initialization of the request the db is open/closed, for the > processing of the request, the db is opened and closed. All of these > are using the one and only C++ class to open and close the db. This > approach allows me to keep all the open's and closes VERY close > together, there might be lots of code that gets called because of a > function call in between, but the function that does the open, does > the close. This is why I took this approach. Since there are > different threads access the DB at the same time, could this approach > be causing problems I am unaware of? Would I be better off switching > over to an approach of: each request opens the DB once and uses the > same connection through the whole request? My point is: I don't do ANYTHING db initialization in start up phase of loading my module. In the documentation for sqlite3_initialize() it says, "The sqlite3_initialize() routine is called internally by many other SQLite interfaces so that an application usually does not need to invoke sqlite3_initialize() directly." So the Apache module never calls sqlite3_initialize() directly, it allows the first sqlite3_open_v2() to call it. I also don't believe any of the reading of module parameters (configuration stuff that happens both times Apache loads the module) every uses the database, they are all reading in info from flat files which are used when a request comes in, to connect to the database. Oh, and this error does *NOT* happen all the time... The first thing the web browser does once it connects is start a slide show, so within 5 seconds there is a new request coming in from each browser all the time. I have to open browser, after browser, sometime it happens on one of the requests from the first browser (normally not) or it happens after opening a bunch of them. The the specific close in which the crash happens is one of two places in the beginning of the request, either in the authentication section or in the process of initialization of the request object, which happens after authentication. I would love to port this code to Linux, but... It is not stand alone, it depends on a .Net WinForms app to do all the initialization and configuration of the Apache web server, there are also a total of 6 different project (one is the Apache module, the rest are Axis2/C Web Services and library classes), there is some Window specific code in the Axis2/C projects because Axis2/C doesn't have good file support. The long and the short of it is, I think it would take me about a month to port to Linux, just to find one bug? Not a good investment of my time. I am hoping to port the whole thing to OSX some time next year (which is why I am using SQLite and not SQL Server Express), hopefully there are better tools there for debugging this type of thing. Sam _______________________________________________ sqlite-users mailing list sqlite-users@sqlite.org http://sqlite.org:8080/cgi-bin/mailman/listinfo/sqlite-users
