I would be surprised if something called SQLITE_INT64_TYPE would work with any type that offers less than 64 bit precision.
Some comments in sqliteInt.h: typedef sqlite_int64 i64; /* 8-byte signed integer */ typedef sqlite_uint64 u64; /* 8-byte unsigned integer */ typedef UINT32_TYPE u32; /* 4-byte unsigned integer */ typedef UINT16_TYPE u16; /* 2-byte unsigned integer */ typedef INT16_TYPE i16; /* 2-byte signed integer */ typedef UINT8_TYPE u8; /* 1-byte unsigned integer */ typedef UINT8_TYPE i8; /* 1-byte signed integer */ A quick check in a few of the .c files turn up 64 bit assumptions in sqlite3VdbeSerialType, sqlite3VdbeSerialGet and anything related to floating point number support. --- Steffen Gutmann <[EMAIL PROTECTED]> wrote: > That said, I am wondering if 32bit-only integers are properly > supported? I downloaded 3.3.8, and executed "mkdir build; cd build; > ../configure; make test" on an (older) Fedora Core 3 i386 Linux > computer and it builds, runs and tests itself (really cool!) without > failing any of the plenty tests. > > I am targeting a custom environment where it would make sense to only > have 32bit integers (no large file support, no large data amount, > compiler does not have 64bit integers). So for finding out if sqlite3 > would work in such an environment, I browsed this mailing list, found > message > http://www.mail-archive.com/sqlite-users@sqlite.org/msg17647.html"; > target=_blank > >http://www.mail-archive.com/sqlite-users@sqlite.org/msg17647.html > and > added the following to the compile flags: > > TCC += -DSQLITE_INT64_TYPE=long -DSQLITE_32BIT_ROWID > > However, when running make test, some of the tests fail: > > avtrans-7.2... > Expected: [eace792e90be464d92d865f980dc0b31] > Got: [0c4af064469f09b4508ede5c18338ec2] > ...more avtrans-tests fail here... > > bigfile-1.2... > Error: file is encrypted or is not a database > ..more bigfile-tests fail here... > > bind-3.2... > Expected: [integer integer integer] > Got: [] > bind-4.1... > Expected: [1 1234.1234 1e-05 123456789.0] > Got: [1 2.06495910676928e-267 2.06497421493723e-267 > 2.06498862119476e-267] > btree-1.6...make: *** [test] Segmentation fault > > I haven't looked at what the tests are actually doing (I though I just > ask here...) and my guess is that some of them simply don't work with > 32bit integers. Unfortunately the last one (btree-1.6) seg faults so > without changing the test procedure it is not possible to see the > output of the remaining tests. > > I am wondering if a 32bit version (using -DSQLITE_INT64_TYPE=long > -DSQLITE_32BIT_ROWID) is really supported by the library? ____________________________________________________________________________________ Do you Yahoo!? Everyone is raving about the all-new Yahoo! Mail beta. http://new.mail.yahoo.com ----------------------------------------------------------------------------- To unsubscribe, send email to [EMAIL PROTECTED] -----------------------------------------------------------------------------
