On 10 July 2017 at 00:22, Dan Ackroyd <dan...@basereality.com> wrote:
> Hi, > > I'm passing on an error report from a downstream library that uses > SQLite - https://bugs.php.net/bug.php?id=74883 - without having > tested that it is an accurate bug report. Apologies in advance if it > is a spurious bug report, however I don't currently have an > environment to test it trivially. > > It seems that trying to open a database with the flags > SQLITE3_OPEN_READONLY | SQLITE3_OPEN_CREATE set leads to a very > non-intuitive error message of "out of memory". > I'm not sure the report is accurate. The very first thing that openDatabase() does is check that the flags make sense; if the following condition is true then SQLITE_MISUSE is returned: ((1<<(flags&7)) & 0x46)==0 The flags themselves are defined thusly: #define SQLITE_OPEN_READONLY 0x00000001 #define SQLITE_OPEN_READWRITE 0x00000002 #define SQLITE_OPEN_CREATE 0x00000004 With SQLITE_OPEN_READONLY | SQLITE_OPEN_CREATE, (flags&7) = 5 and thus: 1 << 5 = 32 = 0x20, 0x20 & 0x46 = 0 Looking at the other cases... SQLITE_OPEN_READONLY 1 << 1 = 0x02, 0x02 & 0x46 = 0x02 SQLITE_OPEN_READWRITE 1 << 2 = 0x04, 0x04 & 0x46 = 0x04 SQLITE_OPEN_READONLY | SQLITE_OPEN_READWRITE 1 << 3 = 0x08, 0x08 & 0x46 = 0 SQLITE_OPEN_CREATE 1 << 4 = 0x10, 0x10 & 0x46 = 0 SQLITE_OPEN_CREATE | SQLITE_OPEN_READWRITE 1 << 6 = 0x40, 0x40 & 0x46 = 0x40 SQLITE_OPEN_CREATE | SQLITE_OPEN_READWRITE | SQLITE_OPEN_READONLY 1 << 7 = 0x80, 0x80 & 0x46 = 0 So you should get SQLITE_MISUSE unless flags&7 is one of the three sensible states: SQLITE_OPEN_READONLY SQLITE_OPEN_READWRITE SQLITE_OPEN_READWRITE | SQLITE_OPEN_CREATE I don't think sqlite is giving you SQLITE_NOMEM here. This is based on code analysis of versions 3.8.1 and 3.18.0 (it's identical in both; I haven't checked inbetween versions). -Rowan _______________________________________________ sqlite-users mailing list sqlite-users@mailinglists.sqlite.org http://mailinglists.sqlite.org/cgi-bin/mailman/listinfo/sqlite-users
